Rusbeh Abtahi

• Scope: A human-controlled GenAI orchestration platform for software development, designed to support the path from requirements and architecture through code assistance, test preparation, retrieved project knowledge, and reusable engineering memory.

• Implemented: Agentic GenAI orchestration core with structured prompt processing, stateless JSON-based agents, AI/deterministic context handling, project-based retrieval functionality, multi-stage memory recording/retrieval, SQL-based persistence, structured logging, metric alignment, and AWS deployment.

• In Progress / Vision: GitHub agents for requirements-architecture-code synchronization, benchmarking, agent fine-tuning, auxiliary agent memory, as well as GHOST2 as a future blackboard-based multi-brain design with shared memory and private specialist contexts.

Technical Product Owner and lead developer for BMW Motorrad internal engineering software applications and tool platforms; technical leadership in GenAI-powered engineering software development and model-based embedded function development.

• Core Responsibility: Function and tool development for engine control units and vehicle measurements for the examination of supplier libraries (Bosch, Magneti Marelli) in the areas of injection, ignition, and speed detection.

• Setup and operation of a customer-specific Hardware-in-the-Loop (HiL) setup (Bosch MicroHiL + Lua Scripting) for reproduction and diagnosis of sporadic ECU resets in a motorcycle engine control unit.

Result: Localization of the cause to a mechanical resonance effect in the stepper motor housing.

• Key contribution to the switch from throttle-based to torque-based driver demand control as well as to the development of an application tool for multidimensional parameter calibration.

Result: improved controllability, more stable driving behavior, and approx. 80% less manual calibration time (internal estimate; reported by the calibration team) thanks to the tool.

• Key contribution to a predictive shifting strategy in the vehicle using model-based rapid prototyping (Simulink/Stateflow on dSPACE) with CAN integration and multi-source context (Powertrain/Chassis + Navigation/Radar/Camera), enabling proactive gear selection for upcoming driving situations.

Result: functional prototype and successful live demonstration to executive stakeholders.