Linu Teresa

From research to production

I'm an M.S. Data Science candidate at the University of Maryland with 2+ years as a software engineer at Accenture, where I shipped Java and Spring Boot microservices and REST APIs serving 1,000+ daily users. I'm now an ML research intern at Vitreous State Laboratory in D.C., where I focus on building reproducible Python evaluation pipelines on real scientific data.

My projects push on making LLMs dependable: a hybrid-retrieval RAG pipeline over biomedical literature, a multi-agent orchestration engine that uses Bayesian modeling to cut hallucination loops, and a from-scratch RLHF stack (SFT, reward models, PPO/DPO/GRPO), all running over data infrastructure built with Spark, Airflow, and Dask. I am also first author on an IEEE RASSE 2023 paper. Outside of code, I love to paint and capture photos.

• Agentic & multi-agent systems
• RAG & retrieval
• LLM alignment / RLHF
• Big-data pipelines
• Backend & APIs