At Sprints & BARQ Systems, I contributed to an AI ServiceNow Support Assistant for RAG-based incident support. I configured ServiceNow access and support groups, and built a Qdrant embedding and deduplication pipeline.
I also integrated the backend with Docker Compose and implemented end-to-end tracing. I developed a benchmark harness to evaluate retrieval, refusal correctness, and confidence thresholds.
For my Nexus Auctions graduation project, I built an API-driven hybrid recommender for live and scheduled auctions. It combines collaborative filtering, content-based filtering, and popularity, with dynamic weighting for users with and without history.
I also developed the ARIA RAG chatbot using Sentence-BERT embeddings, ChromaDB retrieval, and Llama 3.2 via Ollama. My other work includes an NLP pipeline comparing classical machine learning and transformer approaches for fake news classification and sentiment analysis.
