DiagAI Technical Brief

Modern vehicle diagnostic trees contain thousands of trouble codes and interrelated freeze-frame parameters. Intermittent faults—such as transient air-fuel excursions or intermittent CAN bus communication drops—frequently elude static decision trees.

DiagAI deploys specialized agents that cross-examine telemetry from distinct engineering perspectives, correlating diagnostic fault codes against high-frequency physical sensor data.

DiagAI divides reasoning across eight dedicated domain agents: MisfireAgent evaluates cylinder combustion anomalies; FuelAgent inspects BSFC, BMEP, and fuel trim; WLTPAgent checks regulatory drive-cycle emissions compliance; DFCAgent decodes OEM-specific UDS fault memory logic; IUPRAgent tracks OBD monitor readiness; and GearAgent, DrivabilityAgent, and StartWarmupAgent analyze transient drivability and thermal stabilization.

Agents evaluate telemetry in parallel, comparing physical evidence against protocol standards to determine fault probability.

DiagAI deeply understands UDS Service $19 (ReadDTCInformation), Service $22 (ReadDataByIdentifier), and freeze-frame structures. It maps proprietary OEM signals to standardized physical channels across legacy DBC databases.

Deployments run fully on-premise using local model runtimes (such as Ollama or LM Studio) with isolated per-OEM knowledge retrieval indices, ensuring proprietary fault logs and calibration files never leave the facility.

DiagAI is in pilot with select powertrain engineering teams. We do not publish accuracy percentages or benchmark scores because comprehensive statistical sample sizes are still being compiled.

DiagAI generates ranked engineering diagnoses and physical verification steps; it does not replace human engineering sign-off for safety-critical automotive systems.