Breathe fire into your data.
Prescriptive performance optimization. Collect and analyze massive amounts of telemetry, ask questions in plain English, and get on-demand recommendations for tuning. Built for racing teams, athletes, and anything with sensors.
First prototype deployment: a NASCAR Cup Series team
The team's database held 451 tables across 13 schemas, including a telemetry table of 372 billion rows: speed, throttle, brake, steering, RPM, gear, accelerometer and gyro, heading, GPS and lap fraction, sampled through every loop on the track. Engineers could query it. Nobody else could.
Breathing Dragon AI put a conversational analyst on top. Ask about lap performance by session, pit road in and out times, fuel usage, weather impact, or a competitor, and get a chart and an executive brief. One model does the analysis. A second model writes it up for the people who make decisions.
- Data
- 372 billion telemetry rows in MotherDuck; 451 tables; laps, pit road, fuel, events, tracks, series
- Query
- Natural language to SQL with retrieval over the real schema, up to three self-corrections, and fixes that become rules after three successes
- Tools
- Lap analysis, weather impact, event discovery, vector search, charting, executive brief formatting
- Safety
- SELECT-only policy, automatic row limits, zero-row anomaly detection, role-based access, audit log
- Security design
- Zero-trust tunnel, WAF, secrets manager, MFA, two-year audit retention, written for IT security review
Where it applies
Anywhere a sensor stream is too big for a spreadsheet and too valuable to leave to one specialist.
Car telemetry, competitor analysis, pit strategy, track and weather effects across a season.
Running and other sports: training load, pacing, recovery and equipment choices from wearable data.
Autonomous and non-autonomous equipment: performance tuning from logs and sensors, with recommendations an operator can act on.
An admin center for agents, tools, pipelines and data sources, plus an MCP server so other agents can use the same tools.