Owned backend services end-to-end, designing and building a FastAPI-based platform leveraging LLMs, RAG, and vector databases to map academic programs, learning goals, and job skills, serving 10+ universities around the globe.
- Built an LLM advisor agent for curriculum coverage and career-transition skill-gap analysis, combining query decomposition, tool use, and user-specific memory to drive multi-step reasoning over complex advisor queries; improved answer completeness by 20% across 100 evaluation cases.
- Implemented a hybrid retrieval pipeline (dense + BM25) with BGE reranking, self-query, and query expansion for the advisor agent's RAG layer; evaluated using RAGAS, improving Recall@20 to 0.92 and faithfulness to 0.94.
- Developed reusable React components and Next.js pages in TypeScript, delivering interactive dashboards and data visualization features for curriculum-skill mapping.
- Deployed and maintained AI-enabled backend systems on AWS using Docker, ECR, EC2, Nginx, and Gunicorn; orchestrated async workflows with Lambda and Step Functions and automated CI/CD with GitHub Actions.
PythonNext.jsTypeScriptRAGAgentic SystemPostgreSQLAWSDockerOpenAI API
