TIM OH

Advanced Micro Devices (AMD)

• >Reduced manual data collection time by 80% by architecting an automated AI report generation pipeline that aggregates unstructured online data via a custom web search MCP and Python extractor.
• >Ensured hallucination-resistant reports with 100% source attribution by engineering a multi-stage LLM processing engine using a distill-and-generate architecture to consolidate 500+ raw user comments.
• >Cut draft creation time to under 15 seconds by optimizing report generation latency, executing parallel asynchronous LLM calls alongside deterministic database carry-forward logic in Node.js/Express.
• >Achieved sub-1.5s inference latency and 98% transcription accuracy on complex GPU terminology by engineering a real-time speech pipeline using int8-quantized WhisperX and SileroVAD.
• >Reduced average reporting time by 50% by automating technical bug reporting, integrating LLM agents with custom prompt chains to transform unstructured voice audio into schema-validated tickets.

Ford Motor Company

• >Ensured 99.9% update reliability across 5 vehicle models by validating Over-The-Air (OTA) firmware update pipelines over Wi-Fi/Cellular networks for the Telematics Control Unit (TCU).
• >Saved developers an estimated 10+ hours weekly in debugging by engineering a heuristic test-scheduling engine in Python that dynamically ranked test cases by historical failure rates to mitigate flaky tests.
• >Isolated and resolved network handoff latency issues by developing an automated regression testing suite in Python for the vehicle's Wi-Fi hotspot and Emergency-call features.
• >Improved wireless module stability by 15% by optimizing data parsing scripts to analyze multi-gigabyte packet captures from vehicle modems and identify dropped Wi-Fi connections.