Fleet operators
Which vehicles need calibration this week? Silent drift creates liability under expanding AEB mandates — but maintenance teams still rely on driver complaints and periodic scans.
YC Fall 2026 · investor brief · live product
Real-time agent SDK & API for sensor correction
NADIR is the infrastructure layer for ADAS reliability — detect extrinsic drift from existing telematics, forecast time-to-CRITICAL, orchestrate calibration workflows, and export signed evidence. Shadow advisory only. Pilot-ready in days, not homologation cycles.
67 API operations · Python + TypeScript SDK · HORIZON 3D · ~193 API tests in CI
Honest positioning: shadow mode, pilot-ready product, no fake customer logos. Every claim below links to a live surface or CI gate.
Shadow advisory only
FMVSS 127 tailwind
Signed evidence graph
make demo-complete CI
Problem
ADAS calibration is treated like a one-time bay event — but sensors drift every day.
Windshield replacements, parking-lot taps, thermal cycling, and vibration move cameras and radar out of spec. Often with no dash light and no driver complaint — until AEB behaves wrong, lane assist nags, or worse. Fleet directors and repair networks lack a continuous, defensible answer to sensor health.
| What exists today | What NADIR adds |
|---|---|
| "Vehicle X is CRITICAL now" | "Vehicle X has elevated probability of CRITICAL in 11 days" — survival curves + regime badges |
| 2D dashboards and CSV exports | 3D digital twin + CRLB uncertainty ellipsoids in HORIZON |
| Manual evidence assembly after incidents | Cryptographically signed evidence bundles with chain of custody |
| Per-vehicle diagnostics in isolation | Fleet-wide prioritization — knapsack intervention at 2,000-vehicle scale |
| Reactive bay calibration only | Continuous shadow monitoring every day before the bay |
Repair & glass networks
Industry estimates suggest a large share of post-repair ADAS calibrations leave shops without documented proof. MSOs need audit-grade QA at network scale, not one-off PDFs.
Telematics platforms
Partners want a "sensor reliability" module to differentiate — without 18 months of perception R&D and homologation cycles.
Insurers & fleet risk
Claims defensibility requires evidence that drift was detected and remediation was scheduled — not reconstructed after an incident.
Directional context, not legal claims: A 0.5° camera misalignment can materially degrade AEB stopping distance (often cited 30–50% in OEM literature). FMVSS 127 and expanding AEB mandates increase fleet liability for ADAS health. NADIR does not certify safety ECUs or replace OEM homologation.
Why now
Regulatory tailwind meets operational gap.
The hardware stack for ADAS is deployed. The software layer for continuous extrinsic health — across fleets, MSOs, and telematics — does not exist at scale.
FMVSS 127
AEB mandate expands fleet ADAS liability
>80%
Directional: vehicles may leave shops without calibration proof
Days
Shadow pilot close time vs 18–36 mo OEM cycles
API-first
Telematics embed without ECU integration
Complementary positioning: Physical calibration bays (e.g. Kinetic-style hubs) win when a vehicle needs a bay. NADIR wins every day before that — continuous monitoring between visits.
Solution
Telemetry in. Corrected fleet out. You keep the audit trail.
NADIR is a software-only platform: watch sensor health continuously (shadow mode), detect statistically meaningful drift, predict time-to-CRITICAL, prioritize under maintenance budget, and produce signed evidence for insurers and repair networks.
Step 01
Connect telemetry
POST batch ingest from telematics, MSO exports, or simulation. Python and TypeScript SDKs handle schema validation, org-scoped API keys, and tier histograms. Partial commit on batch ingest — production contracts, not mock data.
POST /v1/telemetry/ingest/batch
Step 02
Run correction agents
Tri-tier scoring (Pulse → Meridian → Apex), SSE changepoint streams, and fleet-wide knapsack prioritization. Pick algorithm depth per vehicle — like LLM tiers, but for sensor drift inference.
POST /v1/drift-analysis/residual-score
Step 03
Close the loop
Open workflow cases, verify evidence bundles client-side, push HMAC-signed webhooks to your ops stack. Console for operators; HORIZON for strategic accounts; API for embeds.
GET /v1/evidence/.../bundle
| Fleet director question | NADIR answer | Surface |
|---|---|---|
| Which vehicles need service this week? | CRITICAL-first dispatch queue + A4 prioritizer | Console dispatch |
| What if I do nothing for 14 days? | Survival curves + tier exceedance bands | HORIZON Forecast |
| Can I prove we caught and fixed it? | Signed evidence bundles + factor graph replay | Evidence Theater |
What's built
Diligence checklist — every row has a proof link.
We optimize for "click the product" diligence. No placeholder logos, no vaporware modules. Status reflects June 2026 repo state.
| Layer | Status | Description | Proof |
|---|---|---|---|
| Math SDK SE(3), BOCPD, CRLB, A1–A6 |
Shipped | Core correctness in Python SDK — API and UI are thin consumers | Platform overview · CI: make bench-horizon |
| Production API /v1/* · 67 operations |
Shipped | FastAPI backend, org-scoped keys, quotas, audit middleware | OpenAPI reference · ~193 pytest tests |
| Agent SDK Python + TypeScript |
Shipped | Typed clients, schema validation, tier histogram helpers | Developer quickstart |
| API Theater | Live | Guided 6-step ingest → score → tier compare → evidence arc | Run live demo |
| Dispatch Console | Live | Case queue, workflow closeout, Mahalanobis charts, evidence tab | Launch console |
| Calibration Lab | Live | Collision / thermal / vibration → drift timeline (investor Loom path) | Open lab |
| HORIZON 3D 6 modules |
Iterating | Fleet twin, forecast studio, map, evidence theater, what-if console | Command Deck |
| Algorithm tiers Pulse · Meridian · Apex · Echo |
Live | Per-vehicle depth SKUs — latency, math identity, buyer matrix | Tier pages |
| Investor snapshot | Live | Byte-stable JSON — fleet health, HORIZON surfaces, evidence index | Download pack |
| LOI / pilot tooling | Shipped | Generate LOI PDF, pilot hub, micro-fleet success metrics template | LOI builder · Pilot hub |
| Demo briefing pages | Live | FLEET-204 narrative — console dispatch, fleet twin, evidence theater | Dispatch · Twin |
| Pitch deck | Live | Full fundraise narrative with expandable diligence Q&A | Open deck |
Demo credentials:
nadir_demo_public_readonly_v1 (read-only Console + HORIZON) ·
nadir_demo_sandbox_write_v1 (API Theater live ingest) ·
Hero vehicle FLEET-204 threads through all demo tabs.
Architecture
Three composable planes — software-only deployment.
Existing telematics and dashcam data normalize through connector services, process through drift and confidence inference pipelines, and publish as compliance actions and evidence exports. No ECU writes. No ASIL certification required for shadow advisory.
Customer surfaces (Console, HORIZON, Lab, API Theater)
↓
FastAPI backend (api/) — auth, quotas, org scope, /v1/* routes
↓
Python SDK (NADIR_SDK/python/nadir_sdk/) — all math correctness
↓
Persistence (SQLite/Postgres, evidence store, optional S3/R2)
Core wedge
Agent SDK + API
67 operations, OpenAPI 3.1, Python + TypeScript SDK, tiered inference lanes, HMAC webhooks.
Developers → OperatorsDispatch Console
Case queue, workflow closeout, readonly demo keys for investor walkthroughs.
Launch console → Strategic accountsHORIZON 3D
Fleet twin, forecast studio, evidence theater — the board-meeting surface.
Open HORIZON →| HORIZON module | Route | Function |
|---|---|---|
| Command Deck | #/deck |
KPI rail, 14-day forecast counts, live WS feed |
| Fleet Twin | #/twin/{id} |
WebGL hull, SE(3) frames, CRLB ellipsoids, drift arcs |
| Horizon Map | #/map |
Geospatial tier markers, regional risk |
| Forecast Studio | #/forecast |
Survival curves, regime badges |
| Evidence Theater | #/theater |
Factor graph + SE(3) replay |
| What-If Console | #/whatif |
Intervention scenarios, budget preview |
Market
Five intersecting buyers. One correction layer.
NADIR sits at the intersection of fleet ADAS liability, collision/glass MSO QA, telematics differentiation, insurance defensibility, and OEM fleet validation. We lead with operational reliability infrastructure — not "replace Mobileye."
Regional fleets
50–500 vehicles. Silent drift, no weekly prioritization. Entry: $750 micro-fleet pilot.
National fleets
500–5,000 vehicles. Maintenance budget optimization. Fleet subscription + knapsack prioritizer.
Glass / collision MSOs
Post-repair QA, insurer disputes. Repair-network pilot + signed evidence exports.
Telematics platforms
Embed Pulse API — differentiation without 18 months of perception R&D.
Insurers & OEM programs
Claims defensibility, field health telemetry. Apex + Echo depth for forensic and archival.
| Category | Examples | NADIR position |
|---|---|---|
| Physical calibration bays | Kinetic, OEM service tools | Complementary — monitor between bay visits |
| OEM in-vehicle diagnostics | Mobileye, Bosch internal | Fleet-agnostic layer — MSO, insurer, telematics |
| Fleet telematics | Samsara, Geotab | Sensor reliability module via API embed |
| ADAS scan tools | Aftermarket scan / target systems | Continuous + predictive, not one-time static scan |
2026 GTM motion: US Midwest / Southeast regional fleets and MSOs (founder network). Outbound micro-pilot emails → 15-min demo → $750 LOI → 4-week shadow eval → Meridian conversion. Pilot packages exist for Amazon Logistics-class, Safelite-class, and Caliber-class targets — outreach active, no signed logos claimed here.
Business model
Pilot-first pricing — land at $750, expand per vehicle.
Transparent pilot pricing designed to close LOIs fast, then expand algorithm tier depth. Per-vehicle tiers: Pulse ~$1.50/mo · Meridian ~$3.50–5.00/mo · Apex ~$6.50–8.50/mo.
Shadow kickoff
$750
- 50–200 vehicles
- Read-only shadow monitoring
- 2-week signal report
- Under most VP approval thresholds
Paid pilot
$2,800
- 6 weeks · workflow + evidence
- Meridian-class scoring
- LOI-ready metrics pack
- Success: ≥75% CRITICAL detection confirmed
Fleet subscription
$4.5k+
- From 500 vehicles
- Per-vehicle tier upsell
- API + Console + HORIZON
- Evidence graph switching cost
| Pilot stage | Revenue | Duration | Success metric |
|---|---|---|---|
| Micro-fleet kickoff | $750 one-time | Week 0 | LOI signed, fleet manifest ingested |
| Shadow eval | Included | Weeks 1–4 | ≥85% telemetry completeness |
| Conversion | $750/mo micro sub | Month 2+ | CRITICAL detection ≥75% fleet-manager confirmed |
| Expansion | $4,500+/mo fleet | Month 6+ | Tier upsell + HORIZON for strategic accounts |
Revenue narrative: Land with $750 pilots (low procurement friction). Expand per-vehicle tier depth as fleets add modalities and forecast depth. Retain via evidence graph switching cost — historical signed bundles tied to org workflows. Platform: telematics embed = Pulse volume at thin margin, Meridian upsell at integration layer.
Moat
The fleet product is the wedge. The evidence graph is the moat.
Every deployment adds proprietary drift intelligence — how sensors fail by route, climate, vehicle class, repair type, and remediation outcome. That data does not exist in incumbent tooling.
Signed, replayable, queryable evidence compounds with each pilot. Switching costs rise as fleets attach NADIR bundles to work orders, insurer filings, and telematics dashboards. Algorithm tiers (Pulse → Apex) create natural expansion revenue on the same telemetry stream.
5.2d
Pulse median lead-time (tier bench)
~193
API tests in CI
4 tiers
Pulse · Meridian · Apex · Echo
Shadow
Pilot in days, no ECU writes
What NADIR is
- Continuous shadow monitoring from existing telematics
- Statistically meaningful drift detection and forecasting
- Fleet-wide prioritization under maintenance budget
- Signed evidence bundles as audit artifacts
- API-first embed for telematics partners
- Complement to physical calibration bays
What NADIR is not
- Not an ECU write path or in-loop vehicle controller
- Not ISO 26262 / ASIL safety-certified
- Not a replacement for OEM homologation
- Not a legal opinion — evidence is audit-grade, not court guarantees
- Not claiming signed enterprise logos on this page
- Not slideware — every module links to a live URL
Traction
Started Q1 2026 — product live, pilots in motion.
We report milestones honestly. Pre-seed fundraising in progress. No revenue logos claimed without signed LOIs.
Q1 2026
MVP algorithms — KITTI / nuScenes validation, internal sim targets ±0.05° / ±2 mm.
Q2 2026
Calibration Lab + Console live — production deploy, collision default scenario, demo seed script.
Q2 2026
Pilot packages shipped — Amazon, Safelite, Caliber one-pagers + LOI redlines + /pilot hub.
Q2 2026
Pre-seed in progress — $200K–400K target, 4–6 engineers, 500-unit beta path.
Q4 2026
SDK + API GA path — pilot-hardened connectors, evidence exports, developer docs.
2027+
Seed path — $8–15M after traction; Tier-1 OEM roadmap discussions.
Live
Console, HORIZON, API Theater, Calibration Lab
Shipped
LOI builder, tier pages, investor snapshot
Active
Outbound pilot pipeline (Midwest / Southeast)
CI green
make demo-complete · pytest api/tests
Team
University of Michigan founders — Detroit / Ann Arbor ecosystem.
Deep technical roots in perception, embedded systems, and automotive software. Hiring perception + full-stack + embedded engineers on pre-seed close.
Co-Founder & CEO
Dhruv Hegde
Michigan CS + Math. Computer vision, camera calibration, low-level automotive sensor programming. Owns API/math/LOI velocity and investor demo arc.
- Production API architecture and SDK contracts
- Algorithm tier math and bench CI gates
- Pilot LOI generation and fleet outreach
Ann Arbor, MI · team@nadirai.net
Co-Founder & CTO
Sri Balaji
Michigan Computer Engineering. Embedded systems, real-time optimization, sensor fusion architecture. Owns Console, HORIZON surfaces, and operator UX.
- HORIZON 3D command center and WebGL fleet twin
- Dispatch Console and calibration lab surfaces
- Real-time inference pipeline and edge optimization
Ann Arbor, MI · Detroit / Midwest pilot network
Next hires (12 months): Founding perception engineer · Full-stack founding engineer (React/TypeScript/API) · Embedded / optimization engineer.
Live demos
No slideware. Click the product.
Every link below is a real surface wired to production API contracts (shadow mode). Recommended 15-minute investor arc: API Theater → Console dispatch → HORIZON twin → snapshot JSON.
Start here
API Theater
Guided 6-step arc: batch ingest → residual score → tier compare → workflow → evidence bundle → webhook verify. Sandbox write key included.
Run live demo → Board meetingHORIZON 3D
Command Deck KPIs, Fleet Twin with SE(3) frames and CRLB ellipsoids, Forecast Studio survival curves, Evidence Theater replay.
Open HORIZON → Offline diligenceInvestor snapshot
Byte-stable JSON bundling fleet health, all HORIZON read surfaces, evidence index, and investor quickstart URLs. No API required for Loom scripts.
Download pack →Diligence FAQ
Questions investors ask in the first meeting.
Transparent answers. If we don't know, we say so.
Why shadow mode instead of in-loop control?
Shadow mode is a feature, not a limitation. By staying read-only (no ECU writes, not ASIL-certified), NADIR can sell pilots in days, embed in telematics without homologation cycles, and complement — not compete with — OEM safety stacks and bay-based calibration.
How is this different from Kinetic / physical calibration bays?
Complementary, not competitive. Kinetic wins when a vehicle needs a bay. NADIR wins every day before that — continuous monitoring between visits. We route vehicles to bays; we don't replace them.
What's the evidence graph moat in practice?
Every deployment adds proprietary intelligence: how sensors drift by route, climate, vehicle class, repair type, and remediation outcome. That data is signed, replayable, and queryable — and it does not exist in incumbent telematics or scan-tool databases. Switching costs compound as historical bundles attach to work orders and insurer filings.
Do you have paying customers?
We are pilot-ready with live product and active outbound pipeline. We do not claim signed enterprise logos on this page. Micro-fleet pilots at $750 are designed to close fast and convert to Meridian subscription — ask us for current LOI status on a call.
What does the pre-seed round fund?
$200K–400K target at $4–6M pre-money. Allocation: ~40% team, ~30% hardware/manufacturing, ~20% pilot deployments, ~10% infra/ops. 18-month path: first 500-vehicle pilot, 3–5 pilots, seed prep.
How do I verify the tech works?
Three paths: (1) Run API Theater live with sandbox write key — 6 steps, ~5 minutes. (2) Open HORIZON with readonly key — FLEET-204 CRITICAL vehicle. (3) Download the investor snapshot JSON for offline review. CI gates: make bench-horizon, pytest api/tests.
What are the algorithm tiers?
Like LLM providers offer Opus/Sonnet/Haiku, NADIR offers four lanes on the same telemetry:
- Pulse (~$1.50/veh/mo, <200ms) — χ² Mahalanobis + hard tier gates
- Meridian (~$3.50–5.00/veh/mo, 0.5–2s) — Pipeline v2/v3 + A2 survival + regime
- Apex (~$6.50–8.50/veh/mo, 2–8s) — Full A1–A6 + changepoint + CRLB
- Echo (add-on/batch) — Offline replay at Meridian/Apex depth for legal/compliance
30-minute investor walkthrough.
We'll run API Theater live, walk HORIZON with FLEET-204, share the diligence snapshot, and scope a shadow pilot if there's fit.
Or download the offline diligence pack · full pitch deck