Road deaths in India in 2022. A further 4,43,366 people were injured across 4,61,312 reported crashes. Crash risk is concentrated in how people actually drive — and that behaviour goes unmeasured.
Source: MoRTH, Road Accidents in India 2022.
No dongle. No black box. No truck roll. Sakshya reads a phone’s own GPS, accelerometer and gyroscope, and turns the drive into an objective, tamper-resistant, explainable driver-risk score — powering safer drivers, smarter fleets and fairer insurance.
The risk signal for every driver.
Sakshya is the first product from AIABS — building a non-monolithic AGI architecture where each product is a domain-specific AI expert, orchestrated to help manage fleets, gig-worker economies and insurance networks at scale.
±3.1 pts
p78–p84
41 min moving
A dimension the data cannot measure returns nothing and the remaining weights renormalise — the engine says “not measurable” rather than inventing a 100. Illustrative interface; figures are representative, not a live customer trip.
Scores how you drive, explains why, coaches you, rewards you, and calls for help in an emergency.
Live visibility, safety alerts, two-way messaging, analytics and compliance-grade records.
The objective, anti-gaming, explainable risk signal insurers and fleets can price.
Every party in the motor-risk chain is making decisions on a proxy for driving instead of the driving itself.
Road deaths in India in 2022. A further 4,43,366 people were injured across 4,61,312 reported crashes. Crash risk is concentrated in how people actually drive — and that behaviour goes unmeasured.
Source: MoRTH, Road Accidents in India 2022.
Motor’s share of India’s non-life Gross Written Premium. Premiums are set on proxies — age, vehicle, location — not on real driving. Safe drivers subsidise risky ones and insurers carry avoidable loss ratios.
Source: IRDAI Annual Report FY24.
India motor Gross Written Premium, 2025. A large, growing, mispriced premium pool. Usage-based insurance is now permitted — but insurers lack a trustworthy behaviour signal to price it on.
Source: IBEF, 2025.
Operators cannot see risky driving until after an incident, cannot reach a driver safely while they are in motion, and lack clean records for disputes, claims and compliance.
India’s largest vehicle class is the one hardware telematics cannot economically serve at all — dongles and black boxes do not fit a bike. Yet two-wheelers account for a disproportionate share of road fatalities.
Dongles, black boxes and the AIS-140 mandate mean per-vehicle cost, install friction and slow rollout — a poor fit for smartphone-first, high-turnover, gig-heavy markets.
AIS-140 is mandated for all new commercial vehicles in India.
Software-first telematics and safety, built smartphone-first for emerging markets. A dashcam and a GPS tag are optional fidelity layers — never a requirement.
Trip capture is the hard part, and it is done. The phone’s own GPS, accelerometer and gyroscope produce a clean, complete, offline-resilient record of the drive — then the driver is told exactly why they scored what they scored.
±3.1 pts
p78–p84
41 min moving
A dimension the data cannot measure returns nothing and the remaining weights renormalise — the engine says “not measurable” rather than inventing a 100. Illustrative interface; figures are representative, not a live customer trip.
A full web console for the desk, a dedicated Android manager app for the field, and a separate, IP-fenced platform-admin console for enterprise and insurer procurement. One backend behind all three.
Event stream
Illustrative interface with simulated event data. Engineered for hundreds of thousands of drivers per view — the 200k path is built, not yet deployed at that size, and we say so.
Know Your Driving turns raw phone sensor data into an objective, anti-gaming, explainable driver-risk score, delivered to insurers through a consent-scoped partner API. This is the asset insurers and fleets pay for.
// Partner JWT, scope: INSURER_DATA_SHARING
// Consent-gated. Raw location is never served.
{
"uin": "4421••••8807",
"score": 732,
"confidence": { "ci95": [78.9, 85.1] },
"percentile_band": "p78-p84",
"vehicle_class": "TWO_WHEELER",
"breakdown": {
"speed_compliance": { "w": 0.25, "v": 88 },
"braking": { "w": 0.20, "v": 74 },
"cornering": { "w": 0.15, "v": 79 },
"context": { "w": 0.10, "v": null,
"status": "NOT_MEASURABLE" }
},
"weights_renormalised": true,
"exposure_km_90d": 2841,
"consent": { "scope": "SCORE_READ",
"expires": "2026-11-30",
"revocable": true }
}
Representative response shape. The legacy unscoped endpoint was found in audit and removed.
Every driver gets a permanent, platform-wide 12-digit numeric identity (UIN). It follows them across the driver app, every fleet they work for and every insurer quote; businesses get the matching BUIN. This is what makes three products one platform instead of three integrations — and it is why the score accumulates across job changes instead of resetting.
Sakshya’s AI Fleet Manager turns a fleet console into a conversation. Managers ask in natural language; the bot reasons over live fleet data, surfaces what matters, takes safe actions, escalates the rest — and proactively watches for risk before an incident becomes a claim.
Proactive alert: Driver UIN-4421-8807 scored 52/100 on the morning trip — 3 harsh brakes in 22 min on NH-48. Likely fatigue (8.5 hrs driving, slept 4.5 hrs per pattern). Recommended action: pause next assignment + send coaching card. I’ve drafted the message; tap to send.
Show me the trip.
Pulling trip T-884213. Event map: harsh-brake clusters at km 12, 18, 21. Speed variance +28% vs the driver’s 30-day baseline. No overspeed. Weather clear. I’ve flagged the trip for review.
Who else is at risk in the West zone today?
4 drivers flagged. Top 2: UIN-4421 (fatigue, above), UIN-7733 (overspeed streak — 92 in a 60 zone, twice). Other 2 minor. Want me to send coaching nudges to the bottom 2 and draft escalations for the top 2?
Yes, and add the top 2 to this week’s safety review.
Done. Coaching nudges sent. Escalation drafts ready for review. Safety review updated. All actions logged to the audit trail with the deterministic checks behind them.
Safety-critical actions never fire on an LLM alone. The bot can suggest; the rule engine decides. Every step is logged — for the fleet manager, the driver, the insurer and the regulator.
“Show me all drivers who braked harshly in the last hour.” No SQL, no filters, no training.
The bot doesn’t wait to be asked — it surfaces fatigue, overspeed streaks, low-score drivers and SOS events as they happen, with a recommended action attached.
Pull a trip’s timeline, event map, weather and driver history on request. The bot assembles an incident report with a one-paragraph executive summary.
“Move these 3 drivers off the night shift.” “Send a coaching card to anyone below 65 this week.” The bot drafts the action; the manager approves.
“Which zone is getting riskier?” “Compare this month’s score distribution to last.” The bot reads the analytics dashboard so the manager doesn’t have to.
Every read, every recommendation, every action — recorded with the LLM proposal and the deterministic verification behind it. Insurer-grade by design.
Five stages, one capture. Anyone can read a phone’s accelerometer — the work is everything that happens between the sensor and the score.
The phone’s GPS, accelerometer and gyroscope record the drive — battery-adaptive, offline-first, auto-started. Nothing to install in the vehicle.
Orientation recovery reconstructs the vehicle’s true motion frame from gravity, GPS acceleration and gyroscope, so a phone in a pocket or a cup-holder still yields valid physics.
Per-channel anti-gaming checks run before anything is scored: falsified location, passenger-versus-driver, signal-denied, capture completeness.
Six named dimensions with published weights, a two-wheeler-specific matrix, exposure-correct denominators, and abstention where the data will not support a number.
The driver sees why. The fleet sees the evidence and can message, coach or lock out. The insurer sees a priceable signal — never raw location.
Closed-loop safety intelligence: smartphone, dashcam, GPS and IMU sensing feed multimodal AI understanding, which feeds risk prediction, accident prevention, outcome measurement — and then improves the models that started the loop.
Three tiers, one engine. Six load-bearing systems stand between a raw sensor stream and a number an underwriter will accept.
Phone IMU and GPS — works on any device, no hardware. Powers today’s scoring.
Forward video for event corroboration and dispute resolution. Integrated via an OEM partner API — we do not ship hardware.
A dedicated AIS-140-style hardware tag for underwriting and claims-grade evidence. Same engine, higher bar.
Integrate, don’t build. Each tier composes into the same engine — it never rips and replaces. The score stays continuous as a fleet upgrades.
Reconstructs the vehicle’s real motion frame from gravity, GPS-acceleration and gyroscope. A phone loose in a cup-holder or re-oriented mid-trip still yields valid physics.
Versioned, explicit, per-channel checks: falsified location, passenger-vs-driver, signal-denied, capture-completeness. A GPS blackout can no longer certify as a clean drive.
An unmeasurable dimension is withheld from score, badges, journeys and coaching — weights renormalise. Rare, and the credibility of the whole product.
Moving time, not wall-clock. Population baseline and percentile banding with anchor hygiene, so tiny-exposure drivers cannot distort p5/p95.
Partner JWT scoped to INSURER_DATA_SHARING — serves score, history, feature breakdown, trips, portfolio distribution and quote-handoff verification.
OpenStreetMap speed limits cached offline (~36,900 cells, ~12 MB per metro). One-command deploy with auto-rollback, a single paging path, WAF auth-flood penalty box and CI gates.
Safety, compliance and live ops, sold as per-driver SaaS.
Why Sakshya wins
Live map at 100k+ scale, harsh-event alerts, a trip-review trust queue and webhooks into their transport management system. Replaces a dongle per truck with one app install — and drivers keep their UIN when they churn.
Learn moreDriver scoring and rider-safety differentiation.
Why Sakshya wins
The score explains why a driver is risky, not just that they are. Two-way messaging with a driving lockout, and SOS that closes the loop in seconds — rider trust, not just fleet ops.
Learn moreDriver safety, two-wheeler support and churn-proof identity.
Why Sakshya wins
Heavily two-wheeler, high-turnover, smartphone-only — exactly where hardware telematics fails. The two-wheeler weight matrix scores bikes properly, and the score follows the driver to the next gig.
Learn moreFirst-class 2W scoring — uneconomic for hardware.
Why Sakshya wins
Dongles and black boxes do not fit a bike. Sakshya serves the vehicle class hardware telematics cannot: India’s largest, and the most over-represented in fatalities.
Learn moreCompliance plus parent visibility.
Why Sakshya wins
AIS-140 mandates tracking; Sakshya leapfrogs it software-first. Parent-share, an off-duty boundary and failsafe SOS add up to trust, not just compliance.
Learn moreRisk score and UBI enablement via partner API.
Why Sakshya wins
A consent-scoped JWT API serves score, history, feature breakdown, trips, portfolio distribution and quote-handoff verification — without ever touching raw location.
Learn moreFreemium safety, family plans and insurance referrals.
Why Sakshya wins
The funnel that feeds the data moat. Every free driver improves the models; family sharing and failsafe SOS drive adoption beyond any single fleet or insurer.
Learn more“Anyone can read a phone’s accelerometer. Almost no one can turn it into a driver-risk score an insurer will price — with orientation recovery, per-channel anti-gaming, honest abstention, portable consent-bound identity, and production-grade scale.”
The Sakshya moat, in one sentenceA smartphone-first risk dataset — per-channel anti-gaming, consent-bound, keyed to a portable driver identity — is earned trip by trip, not acquired.
Anti-gaming, explainability, honest abstention and DPDP-native consent: four pillars of priceable signal, not one.
Every trip refines the models and the population baseline. The insurer revenue line strengthens each quarter.
Each fleet or insurer link is DPDP-native, scoped, revocable and audit-logged. Consent is the front door, not a checkbox.
A single phone sensor stream feeds the fleet SaaS view and the insurer risk view. Two value paths, one capture, zero extra collection.
The UIN follows the driver across fleets, gigs and churn — the score accumulates, it does not reset.
Driver and operator see the same harsh-event evidence. Disputes resolve on shared truth, not he-said-she-said.
SOS → alert → driving lockout → kin and fleet notified → incident reconstruction. A safety event becomes a verified, replayable record in minutes.
The partner API serves score, history, feature breakdown and quote-handoff verification — scoped JWT, no re-collection.
Designed Road Risk Index (RRI) — the trip stream as a geospatial risk map: a second product for insurers, cities and reinsurers.
Designed Road-quality intelligence — IMU-derived road-condition mapping for insurers, cities and infrastructure bodies.
Five tailwinds converging on smartphone-first driver safety and insurance technology — none of which existed five years ago.
Modern phone sensors and on-device compute make hardware-free telematics viable at scale for the first time. The hardware advantage of dongles and black boxes has collapsed.
Sensor stack standard since ~2020
India’s insurance regulator has formally permitted Pay-As-You-Drive and Pay-How-You-Drive motor products — but insurers lack a trustworthy behaviour signal to price them on.
IRDAI circular, 5 July 2022
Government emergency-response systems and safety mandates create tailwinds for driver-safety tech — and a national SOS endpoint to plug into.
India 112 ERSS operational nationwide
India’s DPDP Act makes consent-first, revocable, data-minimising design a legal requirement. Sakshya already is; most incumbents are not.
Digital Personal Data Protection Act, 2023
High-turnover, smartphone-only, heavily two-wheeler driver bases are exactly where hardware telematics fails and Sakshya wins. India commercial telematics: $1.97B (2025) → $7.14B (2034), 15.4% CAGR.
IMARC Group, 2025
| Axis | Hardware telematics (dongle / AIS-140) |
Global UBI apps | Sakshya |
|---|---|---|---|
| Hardware needed | Yes — per vehicle | No | Optional — smartphone-primary |
| Onboarding speed | Slow — install | Fast | Fast — install the app |
| Two-wheeler support | Uneconomic | Rare | First-class — own weight matrix |
| Emerging-market fit | Poor — cost and logistics | Weak — localisation | Built for it |
| Consent / DPDP-native | Rarely | Varies | Yes — scoped, expiring, revocable |
| Anti-gaming | Varies | Varies | Explicit, versioned, per-channel |
| Says “not measurable” | No | No | Yes — abstains and renormalises |
| Portable driver identity | No | No | Yes — UIN across fleets and insurers |
| Driver-visible access log | No | No | Yes |
| Failsafe emergency SOS | Rare | Rare | Yes — live public broadcast link |
| Fleet + insurer + D2C in one | No | Rarely | Yes — one capture |
Comparison reflects the general market categories, not any single named vendor.
Telematics fails on trust before it fails on technology. Every control below is shipped in production and enforced in continuous integration — because an insurer’s procurement team will ask.
Granular, revocable, expiring consent for data, live location and chat — one unified predicate across the whole codebase, CI-enforced so no surface can quietly check consent differently.
The driver sees every read of their data. DPDP-grade export and right-to-erasure cascading across 37 tables, with a CI gate that fails the build if a new table escapes the sweep.
Device integrity attestation, biometric sign-in, encrypted local storage, and a signed fail-closed remote-config channel (Ed25519) — nobody can push a malicious config to the fleet.
An off-duty driver is not visible to their employer. Family sharing is entirely separate from any employer. The public SOS link is served from the app domain, never the API domain.
Straight answers, including on the things we have not proved yet.
Ask us something elseNo. Sakshya is smartphone-first: it reads the phone's own GPS, accelerometer and gyroscope, so a driver is scored the moment they install the app and a fleet can onboard tens of thousands of drivers in days rather than quarters. A dashcam and a dedicated GPS tag are optional fidelity tiers that compose into the same scoring engine for underwriting- and claims-grade evidence; they are never a requirement.
Through orientation recovery and multi-source sensor fusion. Sakshya reconstructs the vehicle's real motion frame from gravity, GPS-derived acceleration and the gyroscope, so the phone's own position and orientation - including a change mid-trip - do not corrupt the physics. This is the make-or-break problem of smartphone telematics and it is solved in production.
Explicit, versioned, per-channel anti-gaming checks run before anything is scored: falsified location, passenger-versus-driver detection, signal-denied conditions and capture completeness. A GPS blackout can no longer certify as a clean drive. Where the data genuinely cannot support a dimension, the engine abstains and renormalises the remaining weights rather than inventing a number.
The score is built from six named dimensions with published weights: speed compliance 25%, braking 20%, acceleration 15%, cornering 15%, smoothness 15% and context (time of day and road) 10%. Two-wheelers and four-wheelers use separate weight matrices. Every score carries a per-trip impact breakdown, a confidence interval, a population percentile band, a trip timeline and an event map - so the driver sees why, and the insurer sees a feature breakdown rather than a black box.
No. The insurer partner API is served through a JWT scoped to insurer data sharing and returns score, history, feature breakdown, trips, portfolio distribution and quote-handoff verification - without exposing raw location. Every link is consent-gated, scoped, expiring, revocable and audit-logged, and the driver can see every read of their data in their own access history.
Sakshya was designed consent-first rather than retrofitted. Consent is granular, revocable and expiring, expressed through a single unified predicate across the codebase and enforced in continuous integration. DPDP-grade data export and right-to-erasure cascade across 37 tables with a CI gate that fails the build if a new table escapes the sweep, and every read of a driver's data is logged and visible to that driver.
No. Sakshya is a driver-safety and risk-analytics platform, not an insurer, broker or intermediary. It does not sell, underwrite, solicit or advise on insurance. Its role is to supply a priceable, explainable driver-risk signal to licensed insurers who make their own underwriting decisions.
The Android driver app, the fleet console, the Android manager app, the platform-admin console and the KYD scoring engine are live in production. The insurer partner API and journeys/chaptering are built and awaiting flag-off. iOS is in progress, the live map is engineered for 200,000+ drivers per view but not yet deployed at that size, the Road Risk Index and road-quality intelligence are designed, and score-to-claims validation has not started. Every capability on this site carries a Live, Built, Designed or Planned label for exactly this reason.
Fleet operators and gig platforms can pilot in days — there is nothing to install in the vehicle. Insurers can start with a scoped, consent-gated read of the score. Investors can request the data room.
No cookies, no trackers, no third-party scripts on this site. Your enquiry reaches the founders directly.