coco User

I'm building AI systems that move from model training and evaluation to usable APIs and interfaces. My work focuses on misinformation detection, retrieval-augmented generation, and efficient LLM inference.

For a misinformation risk detection system, I fine-tuned a BERT-based Transformer end-to-end and compared it with TF-IDF and logistic-regression baselines using F1-score. I exposed the model through FastAPI and analyzed inference logs in Databricks and SQL.

I've also researched LLM cost optimization through routing and clustering, training Random Forest, XGBoost, and MLP models against held-out benchmarks. In a RAG document chatbot, I built document ingestion, embeddings, vector search, and structured prompting while testing chunking and retrieval approaches to reduce hallucinations.

Alongside AI work, I built a real-time color-based object removal pipeline with OpenCV and helped organize a 100+ participant women's hackathon. I'm pursuing a B.Tech in Computer Science & Engineering with an AI/ML specialization at IIIT Nagpur.