SNEHA KR

Researched and optimized unsupervised algorithms in PyTorch to improve source signal separation quality by 50%, surpassing existing benchmarks. Collaborated with CNRS@CREATE researchers to debug and integrate codebases, accelerating model development and reproducibility. Conducted 12 tutorials for over 300 students, leading discussions on AI ethics, fairness, trust and safety, and guided ML model building for applications in finance, medicine, and environment.

Researched and optimized unsupervised algorithms in PyTorch to improve signal reconstruction quality by 50%, surpassing benchmarks. Collaborated with CNRS@CREATE researchers to debug and integrate codebases, enhancing development and reproducibility. Conducted tutorials for over 300 students on AI ethics, fairness, trust, safety, and practical ML applications in finance, medicine, and environment.

Developed a generative AI tool using LLMs (GPT, Llama), prompt engineering, and retrieval augmented generation to automate synthesis and Q&A of unstructured pharmaceutical data, reducing APAC research time by 90%. Enhanced semantic retrieval via LLM-driven query expansion and multi-vector retrieval for improved document relevance, accelerating decision support. Applied NLP with Python, Databricks, Spark, and AWS to analyze social media discourse for disease insights supporting product-market fit. Communicated findings to cross-functional stakeholders.