I build long-running AI systems that act, reason, and adapt. Current focus: Autonomous Agents that sift through noisy data to take action with minimal human oversight. I enjoy tightening feedback loops, designing memory systems, and building infrastructure that scales beyond the demo.
| Languages | AI & Data | Infrastructure | Web |
|---|---|---|---|
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
 |
|||
 |
Autonomous Resume Tailoring Agent
End-to-end resume customization aligned to specific job descriptions.
- Problem: Manual resume tailoring is repetitive, error-prone, and often misses role-specific nuance.
- Solution: A multi-step agent workflow that extracts requirements, analyzes tone, and rewrites bullets to balance keyword alignment with narrative consistency.
- Outcome: Drastically reduced application time while improving ATS screening performance.
- Stack:
Crew AIMongoDBGCP
Live F1 Telemetry Analytics
Real-time data bridge connecting F1 telemetry to LLMs for conversational strategy analysis.
- Problem: Live F1 data is fragmented, making it hard to reason about strategy in real-time.
- Solution: An MCP server streaming live telemetry via WebSockets, filtering for signals like tire wear and pit windows before feeding them to an LLM.
- Outcome: Transforms raw streams into an interactive, natural-language strategy interface.
- Stack:
FastMCPHttpXWebSockets
Hybrid-Search RAG Assistant
Precision answer extraction for large technical documents.
- Problem: Keyword search misses context; semantic search misses exact technical details.
- Solution: A hybrid pipeline combining vector similarity with keyword filtering to cite specific sources in large docs.
- Outcome: High-confidence retrieval for technical documentation.
- Stack:
TypeScriptNext.jsPostgreSQLPinecone
RL Traffic Control
Dynamic traffic light optimization using Reinforcement Learning.
- Problem: Fixed-timer intersections are inefficient for variable traffic patterns.
- Solution: A SARSA-based agent that learns to switch lights based on real-time queue length observations.
- Outcome: Reduced average wait times compared to static schedules.
- Stack:
PythonNumPyPygame