Sensor drift, stopped at fleet scale.

NADIR is a software-first calibration intelligence layer: we ingest existing vehicle telemetry in shadow mode, score extrinsic drift and residual confidence, and produce severity-ranked events plus signed evidence bundles for insurers, fleets, and OEM audit.

We do not require you to replace perception or safety ECUs in pilots — connectors + SDK + cloud normalization ship first; optional hardware attach is a later OEM channel play, not a prerequisite for revenue.

• Typical pilot: 250–500 vehicles, 6 months, one data connector + NADIR SDK.
• Outputs: drift tiers, intervention timelines, REST exports, simulation replay fixtures.
• Mean time to detect target: 4.8 minutes on reference fleet replay.

Bay and factory calibrations are snapshots. NADIR is a continuous observability layer — windshield swaps, curb strikes, temperature cycles, and vibration all move extrinsics after the last procedure.

On public driving datasets (KITTI / nuScenes replay), our MVP targets ±0.05° angular and ±2 mm translation residual bands in controlled simulation — with Mahalanobis gating at 99.7% confidence before an alert fires.

Near-term: fleet SaaS ($8–15/mo/vehicle) and repair-network compliance programs. Mid-term: insurer data API ($2–5/mo/vehicle) once evidence volume scales. Long-term: OEM/Tier-1 license ($45–75/vehicle) when hardware programs attach — we model software ARR leading hardware by 24+ months.

We sell to fleet ops, safety/compliance leads, MSO technical directors, and claims innovation teams — buyers who feel downtime and liability, not only R&D engineering.

MVP drift engine and SDK paths are live in shadow mode Q1 2026. Calibration Lab, fleet console, visuals lab, and pilot packages (Amazon / Safelite / Caliber) ship on nadirai.net — synthetic + recorded telemetry without custom silicon.

Fleet LOI in progress (250-vehicle regional delivery pilot). Pre-seed $200K–400K at $4M–6M pre-money. Hardware alpha remains on the roadmap for OEM co-design, not current pilots.

One drift graph feeds multiple SKUs: fleet compliance SaaS, shop validation subscriptions, per-export evidence fees, insurer risk segments, OEM validation APIs, and simulation licensing for digital twin teams.

See Economics for layer-by-layer pricing and margin assumptions.

SDK (C++/Python): ingest adapters for camera/radar/IMU/CAN topics, temporal alignment bus, residual scoring hooks, and local alert thresholds. Ships with synthetic fixtures for <90s first replay in dev environments.

REST API: fleet drift tiers, vehicle-level health, intervention dispatch, evidence bundle retrieval, and simulation-oriented export formats for digital-twin teams.

Pilots typically integrate one connector (e.g. fleet telematics bridge or rosbag/MCAP path) — we do not require new ECU programs to begin shadow mode.

Core pipeline: SE(3) extrinsic estimation with sliding-window bundle adjustment; camera–radar–IMU fusion via EKF/UKF; changepoint detection on residual streams; Mahalanobis + χ² gates at 99.7% confidence before escalating severity.

Implementation stack: C++17/20 numerics (Eigen, Ceres, GTSAM), optional CUDA/TensorRT paths for learned residual scorers, OpenCV geometry primitives. Validated on KITTI / nuScenes replays — MVP targets ±0.05° / ±2 mm in simulation harnesses.

Edge (when deployed): time sync, residual computation, gating, and buffered evidence frames — sized to fit existing fleet gateways or development rigs, not a bespoke NADIR box in current pilots.

Cloud: fleet normalization, route/region segment risk, liability timeline assembly, insurer/OEM export formatting, and dashboard surfaces. This is where most pilot integrations start — upload or stream, then score.

Shadow-mode only: read-only ingestion on approved buses/topics, canonical event stream, no actuation. Dispatch outputs are recommendations and work orders to human technicians — not closed-loop braking/steering overrides.

Designed for AUTOSAR-adjacent fleets and repair networks that cannot recertify safety ECUs per software iteration.

Reference designs (Orin-class compute, STARVIS-class camera, BMI270-class IMU) inform OEM conversations — planned alpha Q4 2026. Revenue model treats hardware as attach on Tier-1/OEM programs; software ARR is the near-term wedge.

We model calibration spend outpacing raw vehicle unit growth because ADAS sensor count per vehicle is rising — more cameras, more radars, more procedures per repair event.

• Calibration services CAGR: 12.8% (2033 SAM $8.4B)
• Calibration tools CAGR: 13.1% (2035 SAM $5.0B)
• Calibration-as-a-service CAGR: 3.4% on installed base

Collision repair is the largest line item because post-event procedures are mandatory yet poorly evidenced. Fleet and insurance are faster software buyers — fewer vehicles per contract but higher monitoring ARPU and data attach.

• Collision repair: 42% (~$3.5B) — MSO, dealer, glass channels
• OEM production: 25% (~$2.1B) — EOL + validation programs
• Fleet management: 21% (~$1.8B) — uptime + compliance buyers
• Insurance / warranty: 12% (~$1.0B) — severity + UBI innovation

Regional delivery fleets (250–2,000 vehicles), multi-shop repair networks modernizing ADAS bays, and simulation/OEM validation teams that need drift-labeled datasets — all integrable without hardware swaps.

Static factory calibration and siloed in-house stacks — no cross-OEM aftermarket evidence layer, no continuous fleet normalization, limited insurer-grade exports.

• Continuous SE(3) monitoring vs one-time bay events.
• Vertical coverage: factory → aftermarket → fleet → insurance.
• Automated compliance validation with signed intervention graphs.
• Cross-OEM AUTOSAR-compatible deployment posture.

Shadow-mode dashboards, drift tiers by route/region, dispatch hooks to bay groups, and sample evidence bundles — typically under 90 seconds to first synthetic replay via SDK fixtures.

• CAC: $480 · ARPV: $220 · Gross margin: 82%
• Customer lifetime: 8 years · LTV:CAC target 3:1 by Y2–Y3

• Y1 2025: $0.8M · Y2: $4.2M · Y3: $18.5M · Y4: $52.3M · Y5: $124.7M (conservative case)

Account-level progression: $1.0k → $2.8k → $9.6k → $28k → $66k (Y1–Y5).

Fleets buy Layer 1 first (uptime + safety ops). Repair networks buy Layer 2–3 (bay compliance + evidence). Insurers buy Layer 4 once export volume proves lift. OEM/Tier-1 activate Layer 5 when validation APIs embed in production and service tooling.

We can land multiple layers on one logo over 24 months without separate product builds — the drift graph is shared infrastructure.

Faster procurement, measurable ROI in months, and dense telemetry for algorithm hardening — OEM cycles are 18–36 months; fleet pilots fund engineering and case studies for Tier-1 conversations.

Dhruv Hegde — Co-Founder & CEO. Michigan CS + Math; computer vision, camera calibration, low-level automotive sensor programming.

Srivatsan Balaji — Co-Founder & CTO. Michigan Computer Engineering; embedded systems, real-time optimization, sensor fusion architecture.

• Founding perception engineer
• Full-stack founding engineer (React/TypeScript/API)
• Embedded / optimization engineer

• M1–6: core hires, beta production, first 500-vehicle pilot.
• M7–12: 3–5 pilots, ~3K vehicles, 15 shops, $500K–1M ARR validation.
• M13–18: seed prep ($8–15M), ISO pathway start, $2–3M ARR target.

• 40% team · 30% hardware/manufacturing · 20% pilot deployments · 10% infra/ops

Mitigation: fleet + aftermarket ARR first; Tier-1 channel partnerships; diversified revenue mix before OEM dependence.

Mitigation: cross-OEM neutral platform, IP posture, defensible aftermarket evidence moat, faster iteration vs siloed programs.

Mitigation: Phase 1 on ASIL-B profile for fleet/aftermarket; parallel consultants (TÜV SÜD, SGS); seed budget line for ASIL-D path.

Mitigation: pilot case studies, channel partnerships, insurance co-marketing to fleets.