Sadika Sood

Curated and validated large-scale training datasets for LLMs across Python and SQL reasoning tasks, improving model code-generation accuracy. ● Applied Reinforcement Learning from Human Feedback (RLHF) to refine model reasoning, reduce hallucinations, and improve instruction adherence. ● Evaluated model outputs using structured benchmarking frameworks measuring factual accuracy, logical reasoning, and computational

Designed an automated clinical variant prioritization pipeline in Python to analyze cancer mutation datasets including SNVs (VCF), CNVs, and molecular biomarkers. ● Integrated multiple knowledge bases (ClinVar, VarSome, cBioPortal) to classify variants into ACMG clinical significance tiers. ● Implemented multi-stage filtering logic to accelerate identification of actionable genomic alterations.

Engineered a multi-stage RNAi candidate discovery pipeline, implementing consensus ranking algorithms and strict positional constraints for siRNA/shRNA design. ● Developed a biomedical semantic search system using ontology-driven query expansion and BioBERT embeddings, improving literature retrieval accuracy. ● Authored technical documentation and Standard Operating Procedures (SOPs) for end-to-end discovery workflows.