The systems we depend on are becoming more autonomous, more complex and harder to verify. A dashboard may show success. A sensor may report a convincing position. A compliance report may present a confident number.
But what happens when the underlying evidence tells a different story?
Designed and built in India · प्रणाली को परखें। दावे को परखें।
Navigation that keeps working when GPS is spoofed, test campaigns for drone firmware, and GNSS integrity for airlines. Engagements start as a funded trial or development partnership.
Evidence packs and software for filings, claims and machines, rebuilt from your own records. Engagements start as a fixed-fee assessment or a shadow pilot.
We help teams look beyond the final result — to what actually happened, what the evidence supports, and where uncertainty remains.
Whether the challenge is an autonomous system, a navigation signal or a compliance claim, the goal is the same: decisions based on evidence you can stand behind.
We turn that question into practical checks: testing whether navigation can be trusted, examining what machines actually did, and rebuilding important figures from their underlying records.
SRAVIL builds software and carries out independent assessments to establish whether systems, operational records and important claims can be trusted. You can engage us in three ways:
Runs on your system or data — for example Achyux on a drone, or Tax Path at a CA firm.
We examine your records or system and hand back a report, audit or evidence pack — for example a CBAM pack or a fleet GNSS audit.
Work done on your site — plant energy and steam-system surveys.
Available now = you can commission it today on your own data. Pilot stage = it still needs measuring on your system — the pilot is that measurement. More on how we label offerings →
We apply the same discipline in four areas. Pick the one closest to your problem.
Is the position your aircraft or drone reports actually true?
Explore Aviation →
Where was the ship really — and what does the record prove?
Explore Marine →
Did the machine really do what its log says it did?
Explore Robotics & Infrastructure →
Will your filing or claim hold up when someone checks it?
Explore Compliance & Assurance →
Representative drone — not a Sravil flight trialWhen GPS is spoofed, a drone can report a convincing position that is simply wrong. Achyux checks that position against sensors the drone already carries and, when GPS can't be trusted, keeps it navigating on them — no extra hardware.
So far it has been tested in simulation on real Indian terrain, including flat ground where other methods struggle. Flight trials are the next step, and we say so plainly.
Simulation Explore Achyux →
A robot can briefly cross a safety limit between the moments its log records. In real humanoid robot logs we found 951 such hidden excursions — and certified none of them as safe by mistake.
Real-data How we check robot logs →
Under CBAM (the EU Carbon Border Adjustment Mechanism), if an Indian exporter can't back its emissions figure, the EU buyer falls back on costlier default values. We rebuild the figure from the plant's own records and show what's missing — before the verifier does.
Real-data How we prepare a carbon-border filing →
Everything we hand over says what passed, what failed and what is still uncertain — and how far each result has actually been proven. We agree the pass/fail criteria with you before we start, and our fee is the same whatever the verdict.
Tell us about the system, record or claim you need to assess. We'll help you work out what can be checked, what evidence it needs and what a useful first step looks like.
Discuss a verification needYour aircraft can report a confident position even when GPS is being spoofed. SRAVIL helps you test whether that position can be trusted, using independent evidence the aircraft already carries — and gives you a record a regulator, insurer or flight-safety board can read.
Available now = you can commission it today on your own data. Pilot stage = it still needs measuring on your system — the pilot is that measurement. More on how we label offerings →
Airbus A320 flight deck — representative, not a Sravil productSpoofing and jamming have moved from conflict zones onto everyday routes. A receiver reports a confident position either way, so the only way to know is to check it against something else the aircraft already carries.
DGCA's crew-reporting window for GNSS (satellite navigation, such as GPS) interference since November 2025 — crews now have to say what happened, fast.
GNSS interference cases reported in India over two years — up from 465 near the western border a year earlier (Civil Aviation Ministry).
the year EASA authorities begin reviewing Part-IS (EASA's information-security rule for aviation) compliance — including how an operator knows its GNSS integrity function works.
Hand the regulator, insurer or flight-safety board a measured record of what your systems detect — not a vendor datasheet.
Checks run on sensors you already carry: IRS (inertial reference system), camera, IMU, radar altimeter. Weeks to a pilot, not a re-fit.
Replay your own logs on a sealed verifier and read the verdicts before any contract — no source exposed.
When an incident turns on what a receiver did, you hold a signed, reproducible analysis instead of an opinion.
An airliner crosses a region where GPS interference is common. Its receiver keeps reporting a smooth, confident position — but for several minutes that position is being pushed off course by a fake signal. The crew has ten minutes to report it. Afterwards, the airline needs to know what the receiver actually did, and whether its own checks would have caught it.
Head of Flight Safety and FOQA/FDM managers dealing with spoofed flights and the DGCA crew-reporting SOP.
Specific-category operators who owe SORA (the drone-operations risk assessment) / ASTM F3269 (the run-time assurance standard for uncrewed aircraft) evidence per vehicle and per firmware release.
Teams who want an independent integrity layer, or verification of the one they ship.
An airline needs two things: a warning when the aircraft's position can't be trusted, and evidence of what actually happened afterwards — for its own safety team and for the auditor.
Tells an airline's crew and safety team when the aircraft's GPS position can't be trusted — and, after an incident, what the spoofing actually did.
The problem. Airliners are being spoofed at cruise, and the receiver cannot tell. Crews need to know when GPS is lying, and airlines need to know afterwards what an attack actually did inside their receivers.
If it goes unresolved. Crews act on a position that may be false, and after the flight nobody can say what the receiver did — or show the regulator what was checked.
What we do. An advisory monitor that judges GPS against the inertial reference system the aircraft already carries, plus an RF-observable layer that catches the coherent spoof the inertial check cannot. The same engine runs as forensics on an airline's own incident logs. Fail-passive by construction, with every detection floor stated.
Next step. Send the QAR (quick-access recorder flight data)/FDR and GNSS logs from one spoofed flight. Start here →
An independent test of whether your aircraft's GPS-integrity function really detects spoofing, written up as the record auditors now ask for.
The problem. Part-IS (Reg. (EU) 2023/203 and 2022/1645) makes an information-security management system mandatory for operators, and authorities begin reviewing it during 2026 audits. Spoofing is now a named risk to a safety-critical function — and an auditor asking “how do you know your receiver detects this?” has no good answer today.
If it goes unresolved. The auditor's question — “how do you know your receiver detects this?” — has no evidenced answer in your ISMS.
What we do. We attack the operator's GNSS integrity function with the spoof classes actually seen on its routes, measure whether it detects them, and emit the assessment record Part-IS demands but does not define. Receiver vendors structurally cannot sell this: it measures how badly their own product fails.
Next step. Send your receiver configuration and one route's recorded GNSS data. Start here →
A drone has no pilot on board to notice it is being misled. These offerings detect spoofing, keep navigation going, and produce the safety evidence regulators and customers ask for.
Zero additional hardware. Software that detects GPS spoofing and keeps the drone navigating on the camera, IMU, radar altimeter or lidar it already carries — no new antenna, radio or payload, and no control authority.
Limits. No flight results yet; performance on your airframe is what the pilot measures.
Software that stops a drone from silently following a fake GPS position.
The problem. A UAV (uncrewed aerial vehicle) cannot tell a spoofed GNSS fix from a real one, and the autopilot follows whichever it is given. BVLOS (beyond visual line of sight) operations cannot be allowed to silently follow a spoof.
If it goes unresolved. A spoofed drone keeps flying confidently in the wrong place; on a BVLOS mission there is no pilot watching to notice.
What we do. A ranging-versus-terrain cross-check with inertial coast — detector and navigator — plus an RF-observable monitor proven on real recorded spoof captures. The reported position is checked against the ground underneath the aircraft and rejected when the terrain disagrees.
Need continued navigation, not only detection? See Achyux, the full sensor-fusion product.
Next step. Send your airframe's sensor list and a set of flight logs. Start here →
Simulation test campaigns that give each drone software release the safety evidence regulators and customers ask for.
The problem. Small OEMs need SORA and ASTM F3269 evidence for every firmware release, and most have no test culture to produce it.
If it goes unresolved. Each firmware release ships without the SORA / ASTM F3269 evidence an authority or customer can ask for.
What we do. A campaign per release on a real PX4 (a widely used open-source autopilot) software-in-the-loop runner. Pass/fail bars are locked before the run, and the verdict never moves a bar after the fact.
Next step. Send your next firmware build and mission profile. Start here →
Safety monitors for drone missions, plus a tamper-evident record of every decision they made.
The problem. Run-time assurance is demanded by ASTM F3269 and SORA but built bespoke per vehicle, and flight logs hold no record of what a monitor decided.
If it goes unresolved. After an event, the logs show what the drone did but not what its safety monitor decided — the record an insurer or authority will ask for.
What we do. Safety monitors validated in PX4 software-in-the-loop before anything touches an airframe, with every gated decision written to a hash-chained evidence record built for EU 2019/947 Art. 19 and the insurer's questions after an event.
Next step. Send your vehicle model and a mission envelope. Start here →
Earlier-stage, specialist or enquiry-only work — available, and labelled honestly.
Spoof checks added to a drone navigation system you didn't build — switched on only once your own data shows they work.
Spoof and consistency guards fitted to third-party drone navigation, tested against injected faults, and delivered default-off until the evidence on your data says switch them on.
Limits. Early bench work against injected faults, not yet wired into a PX4 stack; each guard needs calibration on your data before it is switched on.
Most fleets fly a navigation stack someone else built, with no integrity check of their own on it.
Checks a military rotorcraft's reported position against the terrain profile its radar altimeter already measures, using the DTED already on the airframe.
Limits. Counsel-gated; figures shared under the right agreement. SCOMET review precedes any cross-border sale.
Proves a timestamp held to its UTC duty (MiFID II RTS 25, IEEE C37.118) during a spoof — for trading venues and grid PMUs.
Limits. Counsel-gated; figures shared under the right agreement. SCOMET review precedes any cross-border sale.
Send recorded flight logs or RINEX. A sealed verification server replays them and returns verdicts — no source exposed, before any contract.
Pass/fail criteria are written down and agreed before the first run. They never move afterwards.
Scoped 4–8 weeks on your own aircraft or records, in shadow mode — nothing we run touches your live system.
Per-tail or per-fleet licence, or a repeat campaign per firmware release — only if the verdict supports it.
No. The airline line is an advisory, EFB-class annunciator; the drone layers are fail-passive and have no control authority. DO-178C is a design property here, not a certificate we claim.
No. The checks run on sensors you already carry — IRS, camera, IMU, radar altimeter or lidar.
Every offering carries a maturity label. Recorded-RF results are labelled Real-data; SITL (software-in-the-loop simulation) results are labelled Simulation and never quoted as customer results.
Yes. The sealed verification server lets you replay your own logs and read the verdicts before any contract.
They sell through a counsel-gated channel, with SCOMET review before any cross-border sale.
We replay it against our checks before you commit to anything, and tell you plainly what the evidence shows.
Discuss a pilotA vessel's recorded track can be displaced by spoofing — and navigation, operations and any later investigation can turn on where the ship really was. SRAVIL checks tracks against independent evidence in the logs you already keep, and states what the record can and cannot show.
Available now = you can commission it today on your own data. Pilot stage = it still needs measuring on your system — the pilot is that measurement. More on how we label offerings →
Container ship at sea — representative, not a Sravil clientSpoofing displaces AIS (the ship position-reporting system) and GNSS records at scale. Before anyone relies on a track — for navigation, operations or a later investigation — someone has to show which track is real.
ships affected by GPS jamming in the Middle East Gulf by 7 March 2026, as reported by Windward.
slow walk-offs can hide below AIS reporting cadence — so every check we run states its own blind spot.
(the offshore marine contractors' association) has warned that GNSS interference can take out several position references at once.
Audits run on AIS and position logs you already hold.
Every check states the slowest drift it could have missed.
Find out whether "independent" position references would fail together.
Post-mission evidence that an AUV stayed on its line and covered the box.
Owners and managers who need to know whether their fleet's recorded tracks were spoofed or drifted.
DP vessel operators and survey contractors accepting AUV missions.
Are your ships' recorded positions trustworthy? Spoofing and drift leave traces in the logs you already keep. These checks find them — and help show where interference is coming from.
An audit of your fleet's recorded positions that finds where tracks were spoofed or drifted.
The problem. Fleets already hold the logs that would show a spoof or a walk-off — nobody reads them for it.
If it goes unresolved. Spoofed or drifting tracks sit unnoticed in your records until a dispute, inspection or casualty puts them under scrutiny.
What we do. A spoof and walk-off audit over the logs fleets already keep. No hardware, no RF front-end; fail-passive verdicts with the floor stated.
Next step. Send one fleet-day of AIS or position logs. Start here →
Finding where a GPS spoofer is transmitting from — or catching a ship being walked off course — with no new hardware on board.
On the ground, TDOA over fixed timing receivers puts an error ellipse around the spoofer itself. At sea, the same core against gyro, log and radar-fix dead-reckoning catches walk-off with no new hardware on the vessel.
Limits. Both lines are simulation-only. The ground locator needs 4–6 timing-grade receivers and a real capture; the sea line needs a vessel's recorded gyro, log and radar fixes.
Detecting a spoof is half the job; the other half is finding where it comes from, or catching it on a vessel with no new hardware.
A check of whether a vessel's “redundant” position sensors would all fail together under GPS interference.
We measure whether a given vessel's position-reference redundancy actually survives common-mode GNSS interference, starting with a provenance audit of where each reference really comes from.
Limits. The live monitor needs proprietary rig logs we don't yet have; today's deliverable is the provenance audit.
IMCA (the offshore marine contractors' association) has told the offshore industry in print that GNSS interference can take out several position references at once. A vessel's DP FMEA assumes its redundancy survives that; nobody has tested it.
The same evidence principle applies below the surface: a mission record, like a vessel track, is only as good as what can be checked behind it. An underwater vehicle works for hours where no one can watch it. Accepting its survey means proving afterwards where it went and what it covered.
Evidence, after an underwater survey, that the vehicle stayed on its line, and where its coverage has holes.
A software payload that rides on someone else's vehicle — we don't build AUVs. From the mission log it is designed to check whether the navigation solution stayed inside its declared bounds and where the declared area was not covered, using cooperative acoustic ranging and dead-reckoning drift analysis.
Limits. Simulation, real bathymetry and real sonar images only — no vehicle or wet data yet. A reader for real mission logs is not built yet; terrain-aided navigation was ruled out on real Baltic bathymetry; coverage gaps reaching the survey edge need an independent reference.
An AUV runs 8–40 hours fully untethered with no GPS. Survey acceptance turns on proving, after the fact, that the position solution stayed inside its bounds and the survey box was actually covered — and position truth needs USBL/LBL transponders that can cost more than the survey itself.
Barrier transit detection over real bathymetry with cross-confirmation between vehicles, and coverage evidence for mine-countermeasure clearance.
Limits. Counsel-gated; figures shared under the right agreement. SCOMET review precedes any cross-border sale.
AIS or position logs for one fleet and period — read access only.
Detection floors and pass/fail criteria agreed in writing before the run.
Flagged events, a review queue and the detection floor for your data.
Per fleet or per reporting period.
No. The fleet audit runs on recorded AIS and position data; the sea-going spoof check uses the gyro and log the ship already has.
At AIS reporting rates, walk-offs of 0.5 m/s or less can be missed; a ship's own denser position feed restores the full set of checks. Every report states that floor.
We audit it against our checks before you commit to anything.
Discuss a pilotA robot's log can say it stayed within limits when it didn't — samples miss what happens between them. If that record backs a safety case or a regulatory file, the gap is yours to explain. SRAVIL proves what the machine actually did, from the records it already produces, without touching its control.
Available now = you can commission it today on your own data. Pilot stage = it still needs measuring on your system — the pilot is that measurement. More on how we label offerings →
Humanoid robots on public display — representativeRegulation now asks for evidence of what learned and automated machines actually did, and most records can't provide it: logs are sampled, and monitoring often rests on a single instrument.
EU Machinery Regulation becomes mandatory — machines with ML-driven behaviour must keep a safety-relevant behaviour log for five years or more.
is all a learned policy needs to exceed a limit and come back unseen — the most common way a conformance record lies.
— a stuck or drifting piezometer — can hide a developing dam failure or trigger a false alarm.
No model access, no ground truth, no new sensors — verdicts come from existing telemetry.
Shadow-mode engagements run alongside your line or monitoring and touch nothing.
Where the record can't prove something, we state the log rate or extra instrument that would.
Verdict records written for Machinery Regulation and GISTM (the Global Industry Standard on Tailings Management) reviews.
Humanoid and mobile-robot makers who must keep a behaviour log for learned policies under the Machinery Regulation.
Manufacturers and integrators running machine-tending cells with failed cycles and operator interventions.
Owners and engineers of record working to GISTM.
Robots run learned or automated behaviour, and their logs record only snapshots. If a log misses what happened between snapshots, the safety record has a gap you would have to explain — to a regulator, a customer, or after an incident.
Proof, from a robot's own log, that it stayed inside its safety limits — including between the moments the log recorded.
The problem. A humanoid or mobile robot runs a learned policy. Did it stay inside the envelope (its declared safe operating limits) its integrator declared — and can you prove it from the log alone? A log is sampled: a policy can go past a limit and come back inside one sampling interval, so checking samples and reporting “held” is the most common way a conformance record lies. ISO 10218:2025 now covers humanoids but explicitly excludes mobility hazards; ISO/CD 25785-1 is still a draft.
If it goes unresolved. Your conformance record says “held” for moments it cannot actually account for — a gap you would have to explain under the Machinery Regulation or after an incident.
What we do. Every interval in the log gets one of three answers: VIOLATED (a sample is outside the bound), HELD (no excursion was possible between the samples — a proof from a declared slew limit, not an assumption), or INCONCLUSIVE (an excursion was possible and would not have been seen). When it cannot prove, it states the log rate that would have made the interval provable.
Next step. Send one robot log and the limits your integrator declares. Start here →
Software that catches a robot cell's likely failures before the action, so fewer cycles fail and fewer operators have to step in.
The problem. Machine-tending cells lose output to failed cycles and operator interventions, and nobody can say in advance which action was going to fail.
If it goes unresolved. Failed cycles and operator interventions keep eating output, and nobody can say in advance which action is about to fail.
What we do. Runtime assurance for physical action: before a consequential action, the engine determines whether it is executable and safe enough — otherwise it re-observes, defers or abstains, and records why. It runs in shadow mode (running alongside your system without controlling it) against your own line before anything is claimed.
Next step. Send one cell's telemetry and intervention records. Start here →
The same problem exists beyond robots. In critical infrastructure, decisions about safety and structural integrity also depend on whether the instruments providing the evidence can be trusted. Dams and other structures are judged on what their instruments say. One faulty instrument can hide a developing problem — or trigger a false alarm that stops operations.
Spots a faulty dam sensor before it hides a real problem or triggers a false alarm.
A stuck or drifting instrument that alone explains the anomaly is excluded and named. The phreatic surface and factor of safety (the margin before a slope fails) are recomputed without it, and the TARP (trigger-action response plan) trigger is raised on the corrected picture. A fault resting on one instrument stays a WATCH with the reason stated — and the engine says which extra instrument would resolve it.
Limits. Synthetic bench only; not yet connected to any facility's data, and no data-logger connectors yet. Coarse Dupuit section with an infinite-slope factor of safety; internal erosion with no pore-pressure precursor is outside what any instrumented monitor can promise.
A stuck or drifting piezometer can either hide a developing failure or trigger an alarm that isn't real — and the TARP trigger is usually raised on raw readings.
Robot telemetry, cell records or instrument history — read access only.
Your integrator declares the limits; we never fit them from the data. Bars are fixed before the run.
4–8 weeks alongside your line or your monitoring, touching nothing.
A written verdict with what was proved, what was not, and what log rate or instrument would close the gap.
No. Envelope evidence works from the log alone, against limits your integrator declares.
No. Workcell and dam engagements run in shadow mode; nothing we run touches control.
For machines with self-evolving behaviour based on machine learning, a retained log of safety-relevant behaviour — five years or more — from 20 January 2027.
We bench it before you commit, and say what the record can and cannot prove.
Discuss a pilotA filing built on a paraphrase of the rule can cost you a default markup, an unpriced carbon shortfall, an unappealed deduction or a tax election locked in for years. SRAVIL rebuilds the figure from your records and the published rule, and hands back a pack a verifier, payer or chartered accountant can check line by line.
Available now = you can commission it today on your own data. Pilot stage = it still needs measuring on your system — the pilot is that measurement. More on how we label offerings →
Steel coils — the export goods CBAM prices by embedded emissionsEach of these obligations costs money when the figure is wrong or missing — through a default markup, an unpriced carbon shortfall, an unappealed deduction or an election locked in for years. And each is usually prepared from a paraphrase of the rule rather than the rule itself.
what the EU importer applies when an exporter can't show a verified CBAM (the EU Carbon Border Adjustment Mechanism) emissions figure — paid by the exporter either way.
of hospital claim value commonly deducted by payers, and rarely appealed.
the Income-tax Act 2025 renumbered every section — drafts quoting the 1961 Act are quoting repealed law.
A four-week CBAM pack instead of an open-ended consulting project, plus record changes that make next period a re-run.
Avoid default markups, see your CCTS (India's Carbon Credit Trading Scheme) exposure in rupees before the verifier does, and recover deductions the payer's own rules don't allow.
Each figure is recomputed from your records and cited to the regulation, rulebook or statute — quick for a verifier, payer or CA to check.
Quoted per engagement and owed whatever the result; the gap list comes first, so there are no surprises.
Owners facing EU ETS charges, and P&I (protection & indemnity) correspondents and adjusters whose claims turn on a disputed track.
Iron, steel and aluminium producers whose EU customers already ask for emissions data.
~490 entities across nine sectors, and the consultants who prepare their submissions.
Revenue-cycle teams facing payer deductions they rarely have time to appeal.
Practitioners signing multi-year regime elections under the new Income-tax Act.
Exporters, large plants and shipowners now have to back their emissions figures with evidence. We prepare and check that evidence before a verifier, a buyer or a regulator does.
Prepares an Indian steel or aluminium exporter's EU carbon filing, so the EU buyer doesn't fall back on costlier default values.
The problem. Without a verified emissions figure, the EU importer applies CBAM's default markup — a cost the exporter pays either way, through price or lost orders.
If it goes unresolved. Your EU importer applies CBAM's default markup — and you pay it through price or lost orders.
What we do. We build the installation dataset from your own records, close gaps at source on a plant visit, and run the pack against a verifier's evidence checklist so it survives verification.
Next step. Send one plant's records for one reporting period. Start here →
| Line | Source record | Treatment | Status |
|---|---|---|---|
| Natural gas, reheating furnace | Purchase invoices + NCV certificate | Metered value used | Verifier-ready |
| Grid electricity | DISCOM bills, Jan–Mar | Metered kWh × cited emission factor | Verifier-ready |
| Pig iron (precursor) | Supplier declaration missing | Default value applied, flagged | Gap — request from supplier |
| Electrodes consumed | Stores issue register | Process emissions computed | Verifier-ready |
Shows an Indian plant its carbon-credit shortfall in tonnes and rupees before the official verifier does.
The problem. The CCTS compliance year is running. An obligated entity needs to know its emissions intensity, its shortfall in tonnes and rupees, and what the verifier will ask for — before the verifier arrives.
If it goes unresolved. The shortfall surfaces at verification, when the only options left are buying credits or explaining the gap.
What we do. A pre-verification desk check: we recompute greenhouse-gas intensity from your fuel, electricity, process and production records the way the scheme's method describes, compare it with your notified target, and state the surplus or shortfall in tCO2e (tonnes of CO2-equivalent) and rupees, with a “what closes the gap” table.
An integrated steel plant achieves 2.1982 tCO2e per tonne against a notified target of 2.13. Shortfall: 219,139 tCO2e — about ₹26.3 crore at a placeholder credit price of ₹1,200/t. That is the number we put in front of you before the verifier does.
Next step. Send one compliance year's fuel, power and production records. Start here →
What your ships will owe under the EU's emissions trading scheme, worked out before the bill arrives.
The problem. EU ETS (the EU Emissions Trading System) allowance surrender is mandatory for ships calling EU ports, and the exposure has to be computed voyage by voyage.
If it goes unresolved. The EU ETS surrender bill arrives larger than budgeted, voyage by voyage.
What we do. We compute each vessel's EU ETS exposure from its voyage and fuel records, so owners and managers know what they will have to surrender before the bill arrives.
Next step. Send one vessel's voyage and fuel records for a period. Start here →
Claims get decided on evidence that rarely gets checked — a payer's deduction, or a ship's disputed track after a spoofing incident. We check each claim against the published rules and the recorded data, and say how sure the answer is.
Recovers insurance deductions a hospital was entitled to, using the payer's own published rules.
The problem. Payers deduct 10–20% of claim value, commonly, and most deductions are never appealed because checking them by hand costs more than the deduction.
If it goes unresolved. Deductions the payer's own rules don't allow are written off, claim after claim.
What we do. A hospital sends its claims, the payer's remittance advice and its provider contract. We recompute what the payer was obliged to allow on the payer's own published rules, prove where each deduction was wrong, and emit a cited appeal pack per claim line.
On ₹1 crore of claims, a 10–20% deduction rate is ₹10–20 lakh withheld — most of it never appealed.
Next step. Send one batch of claims with the remittance advice. Start here →
Bulk carrier — representative vesselWindward reported GPS jamming affecting more than 1,650 ships in the Middle East Gulf by 7 March 2026. Every casualty and every dispute now turns on attribution: a vessel-originated AIS spoof, external interference, or a genuine track — the distinction that routes the claim under CL.380 (the marine cyber-attack exclusion clause) / LMA 5403. Sravil is building that adjudication from the recorded traces — kinematic plausibility and dead-reckoning consistency, with a co-displacement test across neighbouring ships to follow — and it returns a signed verdict with the uncertainty stated.
An evidence-based verdict on whether a disputed vessel track was spoofed, jammed or genuine.
The problem. Windward reported GPS jamming affecting more than 1,650 ships in the Middle East Gulf by 7 March 2026. Every casualty now turns on attribution: vessel-originated spoof, external interference, or a genuine track — the distinction that routes the claim under CL.380 / LMA 5403 (the marine insurance cyber-exclusion clauses).
If it goes unresolved. A claim stalls or goes the wrong way because no one can show whether the track was spoofed, jammed or genuine.
What we do. We adjudicate from the recorded traces — kinematic plausibility and dead-reckoning consistency, with a co-displacement test across neighbouring ships still being built — and return a signed verdict with the uncertainty stated.
Next step. Send the casualty file — tracks for the vessel and period. Start here →
Some calculations lock you in for years. We reproduce them from the source law, so a professional can sign with confidence.
Ranks a company's long-term tax-regime choices under India's new Income-tax Act, every figure cited — for a chartered accountant to sign.
The problem. The Income-tax Act 2025 replaced the 1961 Act from 1 April 2026 and renumbered every section. An AI draft quoting 115BAA is quoting a repealed statute — and the regime elections it gets wrong lock a manufacturer in for years.
If it goes unresolved. An irrevocable regime election, made on a draft that quotes repealed law, locks the company in for years.
What we do. From a short facts sheet, Tax Path builds the graph of elections the law permits this year, prunes the ones you are not eligible for (citing the section that says so), enumerates every admissible multi-year sequence — with depreciation, loss and MAT-credit carry under option lock — and ranks the best three, every rupee cited to the shipped statute text.
Next step. Send one company's facts sheet. Start here →
On-site work, delivered within existing engagements and labelled for what it is.
An on-site survey that puts a rupee figure on heat lost through a plant's steam pipes and traps.
Insulation and steam-trap losses surveyed on site and priced from physics, delivered within existing engagements by Sravil Field Services. The report is the deliverable.
Limits. Field camera ingestion is still in build; steam-trap acoustics are parked until labelled recordings exist.
Steam-heavy process plants lose energy through insulation and steam traps that nobody has priced.
One plant, one period, one claim batch or one facts sheet is enough to start.
Our first output is the list of what's missing or would be rejected — the agenda, not a surprise at the end.
We rebuild the figure from the published rule and your records, every line cited.
You get the pack and the record changes that make next period a re-run.
No. For CCTS and CBAM we prepare and check before the accredited verifier arrives, and say so on every pack.
No. It is a citable computation for a chartered accountant to sign; the pack says so on every page.
The gap list comes first. Where the rule allows a default, it is used and labelled as a default — never silently.
Per engagement, quoted after scoping, with the fee owed whatever the verdict. Hospital claim recovery is the one recovery-share line, with terms in the contract.
We map it against the original text before you commit to anything.
Discuss a pilotGPS can be manipulated without looking obviously broken — the drone keeps reporting a confident position that is simply wrong. Achyux checks that position against other information the drone already carries, and keeps it navigating when GPS can't be trusted.
No extra hardware. It runs as software on the camera, IMU (inertial measurement unit), radar altimeter or lidar already on board — no new antenna, radio or payload, and no change to the autopilot.
Simulation Software product Pilot stage Bench, PX4 (a widely used open-source autopilot) SITL and real terrain/sensor data — nothing on this page is a flight result.
Representative drone platform — not a Sravil product or flight trialWhen a drone is exposed to GPS spoofing, the reported position may no longer reflect where it actually is. An operator needs to know whether the navigation data can be trusted before relying on it for a mission decision.
Achyux checks GNSS (satellite navigation, such as GPS) against independent onboard sensor evidence, and is designed to support continued navigation when GNSS cannot be trusted — using compatible sensors the drone already carries.
Bench and PX4 software-in-the-loop testing on real terrain data gives encouraging evidence for the approach. These are not flight results; performance on your airframe still needs to be measured.
Most anti-spoof tools stop at the alarm. Achyux checks the GNSS fix against independent onboard witnesses, flags the deception, and — when GNSS can't be trusted — keeps producing a navigation estimate from the same sensors.
TERCOM/TERPROM-class navigation needs relief to match against; on flat ground it falls back to inertial coast. Achyux adds the camera as a witness alongside the IMU. On real near-flat elevation data (Deccan, 2.0 m relief; Thar, 0.8 m) it held 1.2–2.4 m p95 with 98–99% availability.
Drones in a formation each estimate a bearing to the emitter; a fusion node combines them, using bearing and signal strength. Median error falls from about 180 m with three drones to about 110 m with eight.
| What we measured | Result |
|---|---|
| Tested attack classes detected | all tested attack classes |
| False spoof accusations | 0 / 160 attack-free runs on two real elevation lanes, incl. night-thermal (95% upper bound 2.34%) |
| Availability, same runs | 95.5% |
| Position error, nominal | 1.2–1.4 m p95 |
| Worst combined case (night, attack, degraded) | 4.2 m |
Where it was measured. Bench Monte Carlo and PX4/Gazebo SITL on four real elevation lanes (Ladakh, Arunachal, Hyderabad/Deccan, Thar) plus one synthetic lane, with real radar-noise measurements. Acceptance criteria were fixed before every run, and runs are reproducible.
What it does not show. No flight results yet. A procurement-style mountain trial on 22 Oct 2026 added three limits: a drag that ends in a hold under 10–12 m is not detected, by design; the mountain guards slow meacon detection to 6.5–8.3 s; and with GNSS and camera both denied over rugged terrain there is no reliable fix. On steep terrain, radar shadow and map error can disable the terrain witness, leaving navigation on the camera; without the mountain guard, honest jamming false-latched 14 of 30 runs (0 of 30 with it). Attack offsets are synthetic, following publicly reported patterns. Formation localisation is exploratory simulation. Performance on your airframe depends on its sensors and is what a pilot measures.
GNSS interference cases reported in India, Nov 2023–Nov 2025 — up from 465 near the western border a year earlier (Civil Aviation Ministry, Dec 2025).
vessels hit by navigation interference in a single day around the Strait of Hormuz during the 2026 Iran conflict (Windward). A one-day count; by 7 March 2026 Windward reported more than 1,650 ships affected by GPS jamming in the Gulf.
extensive GPS interference documented near Ukraine, the Baltic and the Black Sea.
cost of a seven-day national GNSS disruption, per a UK government study.
| Integration | Measured |
|---|---|
| Interface | ROS2 node, one diagnostics topic |
| Control authority | none — verified every live run |
| Compute | under 1% of budget at 10–50 Hz |
| Memory | 0.31 MiB, no heap after start |
| ARM / aarch64 | bit-for-bit identical |
Your airframe's sensors, rates and compute — what Achyux can use.
Your vehicle in PX4 SITL against the attack suite, bars fixed in advance.
The node on your target compute, with your recorded sensor data.
By agreement — the first rung that would make a figure a flight result.
Achyux is Sravil-owned and sold through a counsel-gated channel; SCOMET review precedes any cross-border sale.
We'll tell you what Achyux can use, and what a pilot would measure.
Discuss a pilotWe help you understand whether a system, calculation or operational record can be trusted — and what to do when it cannot.
Whatever already records the thing in question — logs, records, a filing or a facts sheet. Read access is enough; we never need control of your system.
We read what you send and come back with a gap list: what can be checked, what is missing, and what that means for the result.
No. Checks run on copies of your data, or in shadow mode alongside your system. Nothing we run touches live control.
A written result: what passed, what failed, what remains uncertain and the evidence behind each — with practical next steps.
We state it. Where the evidence can't settle a question, the result says so — and says what extra data would.
You act on the findings yourself, repeat the check on the next release or reporting period, or move to a longer engagement — only if the result supports it.
We check your system or records against agreed requirements and identify errors, inconsistencies and missing evidence.
We show what is missing, what needs correcting and where the evidence may not be enough — before an audit, submission, deployment or operational decision.
A clear report, the supporting evidence, the limitations, and prioritised next steps for the question you need answered.
Where practical, we structure the evidence and records so you can repeat the check, track changes and avoid redoing work.
Our approach: agree what needs to be demonstrated, check it against the relevant evidence, explain the result — and state what remains unverified.
We never move a bar after seeing the result. The verdict is whatever the evidence says against criteria you signed off.
A failing result costs you the same as a passing one, so you can trust that a pass wasn't bought. (Hospital claim recovery is the one recovery-share line, with terms in the contract.)
Runs on your system or data — for example Achyux on a drone, or Tax Path at a CA firm.
We examine your records or system and hand back a report, audit or evidence pack — for example a CBAM pack or a fleet GNSS audit.
Work done on your site — plant energy and steam-system surveys.
Available now means you can commission it today on your own data. Pilot stage means it still needs measuring on your system or real operating conditions — the pilot is that measurement.
Built and specified; not yet measured.
Tested in simulation — sometimes on real terrain, sensor-noise or public data.
Tested on test sets, injected faults or replayed records.
Tested on real recorded operational data.
Proven in live operation with a customer. Nothing of ours is here yet — when it is, we'll say so.
Details shared under the right agreement.
For navigation lines, a sealed verification server lets you replay your own flight logs or RINEX and read the verdicts — no source exposed, before any contract.
What each check does, what you provide, what you receive, headline results with their evidence label, and every known limitation.
An example deliverable, a representative finding, and how pass, fail and unresolved results are reported — on request.
Full test matrices, methodology and implementation detail — shared with technical evaluators for due diligence.
How track evidence decides a marine casualty claim. Read the case on the Marine page →
A trial order, a development partnership with a DRDO lab or DPSU, an iDEX challenge, or a response to an RFI or RFP. A one-page capability statement is available for procurement files.
A fixed-fee assessment or a shadow pilot under a short statement of work, with pass/fail criteria written in before work starts. Software is licensed after the pilot.
Registrations such as DPIIT, GeM and vendor empanelment will be listed on the Trust page once granted.
Scope is public; rates are quoted per engagement after scoping. Machine and navigation lines start as a scoped pilot and repeat per firmware release or per year. A sample anonymised evidence pack is available on request.
Each section names the duty you face, the SRAVIL offerings that answer it, and the usual first step.
GPS interference is now a planning assumption for uncrewed systems. SRAVIL's navigation line is designed and built in India and runs as software on sensors your airframe already carries.
First step. A funded trial: Achyux rides along on an airframe you already fly, publishing diagnostics only, with pass/fail criteria fixed before the flight.
Crews must report GNSS interference within DGCA's 10-minute window, and EASA's Part-IS asks for an information-security record. Both need evidence of what the aircraft's navigation actually did.
First step. Replay one spoofed flight's recorded data and read the verdict.
Customers abroad ask for SORA and ASTM F3269 run-time assurance evidence, per release. SITL campaigns produce it against bars locked before the run.
First step. One SITL campaign on your current firmware.
Ship tracks now decide claims, DP vessels must show independent position references, and EU ETS exposure must be computed per voyage.
First step. Send one fleet day of AIS or telematics.
EU CBAM falls back on costlier default values when an exporter can't back its emissions figure, and India's CCTS sets intensity targets per entity.
First step. One installation's CBAM pack, rebuilt from the plant's own records.
A share of claim value is deducted by payers and rarely appealed. We check each deduction against the payer's own rules (AB PM-JAY, DoH Abu Dhabi).
First step. One claim batch, checked in shadow mode before any invoice.
Since 1 April 2026 the Income-tax Act 2025 renumbered every section. Tax Path computes the election and cites the current section, for the CA to sign.
First step. Send one client's facts sheet.
Robots exported to the EU need behaviour logs from 20 Jan 2027; tailings dams under GISTM are judged on instruments that can drift.
First step. Send one robot log, or one instrument history.
Describe the system, record or filing. We'll tell you which check applies.
Discuss a pilotSRAVIL produces evidence toward these requirements. It is not a certification under any of them. Filing dates are on the Compliance page.
| Rule or duty | What it asks | SRAVIL offering |
|---|---|---|
| DGCA GNSS interference reporting | Crew report within 10 minutes | Airliner GPS-integrity annunciation |
| EASA Part-IS (Reg. 2023/203) | Information-security management in aviation | GNSS integrity assessment |
| SORA · ASTM F3269 | Run-time assurance for drones | SITL campaigns, UAS GNSS integrity |
| EU 2019/947 Art. 19 | Occurrence records for drone operators | Flight monitors & evidence recorder |
| IMCA guidance on DP | Independent position references | Dynamic positioning |
| EU ETS (shipping) | Allowances for voyage emissions | Marine emissions exposure |
| EU CBAM | Verified embedded emissions, or default values | Carbon border filings |
| India CCTS (BEE) | Emission-intensity targets per obligated entity | Carbon credit obligations |
| AB PM-JAY · DoH Abu Dhabi | Payer claim rules | Hospital claim recovery |
| Income-tax Act 2025 | Irrevocable elections, renumbered sections | Tax Path |
| EU Machinery Regulation 2023/1230 | Behaviour logs for ML-driven machines | Learned-policy envelope evidence |
| GISTM | Tailings-dam monitoring and trigger plans | Dam instrument integrity |
| MiFID II RTS 25 · IEEE C37.118 | Timestamps traceable to UTC | GNSS time-transfer integrity |
A two-page brief and a technical annex: architecture, test matrix and limits. Shared on request; detailed matrices under agreement.
An anonymised example of the evidence pack an exporter hands its EU importer.
One page for RFIs and RFPs: what SRAVIL offers, evidence level per offering, and entity details.
Samples are labelled as samples and never presented as customer results. Request a document →
Broadcasting fake satellite signals so a receiver reports a position or time that is wrong but looks convincing.
A spoof that drags the reported position away slowly enough to pass simple checks.
When independent sensors disagree, the position is withheld and flagged instead of averaged.
Software-in-the-loop: the real autopilot software flying in a simulator.
Running beside your system on the same inputs without controlling anything.
Proven within limits · a crossing was observed · the record cannot decide, and the report names what would.
The EU's fallback emission figures, applied when an exporter's own figure can't be verified.
We work on copies of your records or in shadow mode beside your system; nothing we run controls a vehicle, vessel or plant. Retention and deletion terms are written into each engagement.
GNSS, spoofer-location and underwater lines are SRAVIL-owned, developed in India, and sold through a counsel-gated channel. SCOMET review precedes any cross-border sale.
Legal name, CIN, GST and registered office will be listed here before the first contract.
SRAVIL holds no DPIIT recognition, GeM registration, vendor empanelment, ISO or AS9100 certification today. Each will be listed here with its number and date once granted.
When a result doesn't hold up, the site changes.
UAS runtime assurance. "0 false CONTAIN" came from one lucky seed. The scenario generator was fixed and the result now rests on 3 seeds: 18/18 scenarios, 0 false CONTAIN over 699 nominal ticks. A fourth seed re-run on 31 Oct passed 4/6: the drain-margin bar failed on a slow landing estimate, so that bar is not yet stable across seeds.
Airliner GPS integrity. Carry-off detection times updated to the closing measurement: 2 / 1 / 0.5 m/s caught in 27 / 50 / 84 s. An earlier "8 s" figure is withdrawn.
Achyux. A procurement-style mountain trial found three blockers, each closed by a guard package. Its remaining limits are now on the Achyux page.
SRAVIL is the India entity of MASRINT. We are a small team and we say so: engagements start as a scoped pilot with pass/fail criteria written down before any subscription, and a claim we can't substantiate doesn't go on this site.
Pass/fail criteria are agreed in writing first and never moved afterwards.
We're paid for the measurement, not for a favourable answer.
Every result carries its maturity level and what it does not show.
Nothing we run touches your live system.
Spoof detection on recorded GPS signals · a fleet position audit on a real day of ship tracking data · safety-limit proofs on real humanoid robot logs · EU emissions exposure on real ship movement records.
India carbon-credit calculations against the official tables · hospital payer rules · tax computations reviewed by a chartered accountant · drone software test campaigns.
No product is field-validated in live customer operation, and none of our drone navigation work has been flight-tested. Anything marked Pilot stage still needs measuring on a customer's own system — that is what a pilot does.
Every result on this site carries its evidence label, so you can see exactly which of these it rests on.
Legal name, CIN, GST and registered office are being finalised ahead of the first contract and will be stated here once filed. India entity of MASRINT.
GNSS, spoof-location and underwater-surveillance lines are Sravil-owned and sell through a counsel-gated channel; SCOMET review precedes any cross-border sale.
A few lines are enough. We reply with a short scoping note: what can be checked, what evidence it needs, and what a first pilot would look like.
Prefer email? Write to contact@sravil.com.