Projects

Problem: Reading and extracting useful context from research PDFs is slow and difficult to operationalize in downstream AI workflows.

Solution: Built a pipeline that ingests PDFs, performs page or chunk level splitting, stores embeddings with metadata, and retrieves relevant chunks for prompt augmentation.

Impact: Speeds up literature exploration and provides structured context for agentic and LLM-driven research tasks.

Problem: Most interview prep tools generate generic responses and fail to reason over a candidate's actual resume and target role context.

Solution: Built an AI workflow that chunks and embeds uploaded resume and JD documents into ChromaDB, then uses LLM prompting for three features: context-aware Ask, candidate-style Interview simulation, and Improvements analysis with match score and rewrite guidance.

Impact: Improves interview readiness with personalized AI feedback, grounded responses, and higher-signal resume optimization for specific job applications.