
AI Engineer
Intel Corporation · Oct 2023 – Present
- GenAI Agentic Debug Tool: Architected a multi-agent agentic harness (GPT-4 + Microsoft AutoGen, 5 agents: RAG, RCA, image/text analysis, summarizer) delivering net-new GenAI root-cause analysis at >80% accuracy; adopted by ~100 engineers across 4 teams (BIOS, SysDebug, integration, validation), cutting a ~500-ticket backlog by ~20%/week.
- Deployed the GPT-4 backend on Azure OpenAI in a secure hybrid-cloud topology (VNet, private endpoints, internal DNS); built the RAG pipeline over debug logs/BKMs and a RESTful server-client architecture integrating JIRA/HSDE APIs and Jenkins CI.
- Copilot Platform Debug Agent: Designed and shipped an inner-source GitHub Copilot agent + skills library delivering AI-powered root-cause analysis via a custom agentic harness that orchestrates 4 MCP server integrations (HSDE, JIRA, Wiki, co-design) for autonomous ticket fetching, similarity search, and BKM lookup across Intel 10 platform families.
- Cut the 10-engineer SysDebug team's platform-ticket backlog by 35%; built a curated BKM knowledge base (automated OneNote/wiki -> Markdown pipeline) with single-command, zero-admin onboarding.
- Smart IFWI (Integrated Firmware Image) Regression Engine: Sole owner of full lifecycle architected a Python tool (GUI + CLI) for firmware regression isolation with a Claude AI advisor correlating defect-ticket context with binary diffs to surface root-cause components.
- Reduced investigation time from ~109 min to ~6 min (100+ engineering hours/quarter); shipped as a zero-dependency Windows .exe with concurrent HSDE/Jira ingestion via MCP, adopted across 5 teams (50+ engineers).
- Automated PythonSV Installer: Windows tool provisioning Intel PythonSV debug environments across 6 platforms; cut setup from ~30 to ~5 min (83%) with a two-tier Claude recovery pipeline, deployed organization-wide.
- Deep Learning Log Anomaly Detection: Trained a CNN-LSTM model at 99.19% accuracy on an imbalanced dataset for first-level log triage, saving the SysDebug team ~8 engineering hours/day; auto-generates reports with linked HSDE/Jira defect IDs.
- Deployed as a Dockerized Flask app behind NGINX/SSL with MongoDB and Docker CI/CD.


