Aviation AI Pro

A Product by SOS Services

Flight data analytics, built for the safety officer.

Aviation AI Pro ingests ARINC 717 FDR streams, decodes vendor-specific DLU formats, scores every approach against configurable thresholds, and reconstructs flights in 3D — with DGCA/DCDCA-ready exports baked in.

Request a demo See the platform →

ARINC 717 · multi-vendor DLU · DGCA/DCDCA workflows · India-first

What's inside Aviation AI Pro

Eight modules that turn raw flight recorder bytes into safety decisions.

ARINC 717 Parser

ingestion

Industrial-strength FDR decoding, written in C++17 for speed

  • • Fixed, variable, and superframe formats
  • • NRZ, BCD, Discrete, SSM, Manchester encodings
  • • Sync-word sequence detection with configurable patterns
Read more →

DLU Decoder

ingestion

Vendor-specific decoders that don't pretend the world is standard

  • • Honeywell DLU empirical decoder (production)
  • • Per-channel decoders mapped to recorder documentation
  • • Multi-vendor architecture (Honeywell first, more to come)
Read more →

Streaming Ingestion

ingestion

Multi-GB FDR files, bounded memory, no batch jobs left for tomorrow

  • • Memory-bounded streaming with configurable batch sizes
  • • Direct PostgreSQL COPY via libpqxx for high throughput (optional)
  • • HTTP/PostgREST fallback with exponential-backoff retries
Read more →

Approach Scoring

analytics

Detect unstable approaches before the post-flight review meeting

  • • Configurable thresholds per fleet + per airport
  • • Sink rate, glideslope deviation, flap_angle_deg, ILS capture
  • • Per-flight score + drill-down to the second
Read more →

Validation & QA

analytics

Bad data caught at ingest, not at the safety review

  • • CRC validation on every frame
  • • Range validation per parameter (configurable)
  • • NCD (No Computed Data) and test-value preservation
Read more →

PFD + 3D Replay

visualization

See the flight, don't just read the rows

  • • Primary Flight Display (PFD) recreated from FDR parameters
  • • 3D ground track with India airport context
  • • Scrubbable timeline — jump to the exceedance moment
Read more →

DGCA/DCDCA Workflows

compliance

Built for the Indian regulator from day one, not retrofitted

  • • DGCA/DCDCA-aligned exceedance reports
  • • Evidence bundle exports (parsed parameters + raw frames + audit log)
  • • Submission-ready PDF + machine-readable CSV/JSON in one click
Read more →

Audit Trail

compliance

Every parameter has provenance, every change is logged

  • • Per-parameter provenance — frame, slot, decoder version, decode timestamp
  • • User actions logged (threshold edits, manual reviews, approvals)
  • • Immutable audit trail with PostgreSQL append-only enforcement
Read more →

Why Aviation AI Pro instead of a legacy FDR vendor

Modern engineering, not 1990s ports
C++17 parser, FastAPI backend, Vite frontend. No COBOL workflow editors. No 18-month vendor implementation.
India-first regulator design
DGCA/DCDCA workflows are first-class, not bolted on. Submission bundles ship with parsed data, raw frames, and full provenance chain.
Multi-vendor DLU architecture
Honeywell first. Adding the next vendor means adding decoder configs, not rebuilding the platform.
Provenance over claims
Every parameter traces back to its frame, slot, and decoder version. Audit evidence ships with the data, not as a separate PDF.
Per-fleet thresholds
Mumbai-Delhi is not the same approach as Leh. Approach scoring is configurable per airport, per fleet, per crew profile.
Open schema
Time-series Postgres on Supabase. Your data is yours, queryable directly, exportable without vendor lock-in.

Pricing

Per-fleet, transparent, soon.

We're in closed pilot, finishing DCDCA submission readiness. Pricing tiers publish here once we've validated them with regulator + first carrier. For early access, get in touch.

Request early access