CareMind: clinical document QA
Built multi-step query routing, document search, report comparison, and timeline extraction over synthetic medical PDFs. Connected retrieved evidence to cited answers.
LangGraph · FastAPI · Supabase · Next.js
Shree Varaa Mangai V
AI/ML Engineering | Research | Full-Stack Engineering | Computer Vision
M.S. Artificial Intelligence Systems · University of Florida
I build AI applications and evaluate how models fail.
I’m Shree, a recent M.S. graduate and Machine Learning Researcher at the University of Florida. I specialize in benchmarking generative models, building RAG applications, and full-stack engineering, with experience across UF, AGIS, Ford Motor Company, and Spacescan.
Seeking full-time opportunities as an AI/ML Engineer, Computer Vision Engineer, or Software Engineer across the U.S.
15,000+
videos generated for AI evaluation
4,201
prompt–video pairs in the research corpus
94%
PDF question extraction accuracy in Cognito-Map
Selected work
All projectsDetecting when generated videos miss the prompt
Contributed to a 4,201-pair evaluation corpus and TRACE, a detector that uses a frozen vision-language model to identify unsupported prompt content. TRACE-L reached 0.738 AUROC on T2V-CompBench-Wan.
Cognito-Map: PDF extraction and analytics
Built a PDF-to-MongoDB pipeline with Gemini-based question extraction, achieving 94% extraction accuracy, and a dashboard for exploring questions by cognitive level.