01Project records

Inspectable engineering reports.

Each record lists the problem, the measured result, the stack and a sample of what the API actually returns.

Schematic illustration of a lithium-ion battery module with thermal gradient contours
REC.01
live system sample nominal
// POST /api/v1/telemetry/evaluate
{
  "tr_risk_score": 0.0012,
  "soh_percentage": 98.42,
  "latency_ms": 2.28,
  "alert_state": "GREEN",
}
Formula Student EVDual ML API

APEX.AI

Battery Intelligence System · EBISM Backend

A real-time lithium-ion thermal and health engine built for Formula Student EV battery packs. The backend combines safety classification and state-of-health tracking so telemetry can be evaluated inside a sub-5ms runtime budget.

P99 latency

~2.3 ms

runtime

TR false negatives

0.0%

90 tests

SoH RMSE

1.158%

TCN model

  • Paired an XGBoost thermal-runaway classifier with a PyTorch Temporal Convolutional Network exported to ONNX for state-of-health tracking.
  • Mapped API outputs into GREEN, AMBER, RED and ABORT safety tiers with API-key authentication and rate limiting.
  • Connected the service to a real-time CAN bus ingestion flow and completed 19 integration tests.
Python 3.11FastAPIXGBoostPyTorch TCNONNX RuntimeCAN Bus
Airfoil cross-section with laminar streamlines and pressure contours
REC.02
live system sample nominal
// POST /api/v1/airfoil/predict
{
  "cl": 0.842,
  "cd": 0.0187,
  "cm": -0.091,
  "ld_ratio": 45.02,
}
FSAE AerodynamicsCFD Surrogate Model

AERO.AI

Aerodynamic Coefficient Prediction Engine · MVP1 Backend

A lightweight ML surrogate that replaces compute-heavy CFD iterations for FSAE 2D airfoil analysis. The FastAPI service predicts coefficient, stall-margin and lift-to-drag outputs for NACA four-digit airfoils.

Model weight

~5.4 MB

PyTorch binary

Reynolds range

≤ 2.0M

flow condition

Geometry

NACA 4

2D airfoil

  • Produces Cl, Cd, Cm, stall-margin and L/D balance estimates without launching a new CFD solve.
  • Trained against OpenFOAM RANS CFD and XFoil simulation data for engineering-oriented surrogate inference.
  • Uses structured Pydantic schemas and Cloudflare Tunnels to expose remote API access from a WSL2 / Conda workflow.
Python 3.11PyTorchFastAPIOpenFOAMXFoilCloudflare Tunnels