Built AI-powered e-commerce platform. Created classification/data enrichment pipeline that reduced preprocessing costs by 75% compared to naive structured output LLM inference. Engineered ML systems and shipped semantic search + recommendation engine to production.

Designed curriculum and mentored 20 students (2 cohorts, underrepresented focus) building industry-ready portfolio projects (Python); top feedback among 50 mentors with internship placements.

Built AI copilot VS Code extension (Python, Typescript, React) for data analysts using dbt from concept to MVP

• Reduced model dev and review time and minimized cross-team variance through iterative primary key identification (LLM + SQL verification loops), automated documentation generation, and AI-powered data quality checks

• Architected full-stack extension optimizing for cost and latency through multi-layer caching, lazy inference and lazy warehouse querying, and local dbt compilation with DuckDB throwaway profiles

Architected analytics infra (Python, Snowflake, dbt) for top-5 US investment firm on billions of consumer transactions

• Built unified analytics framework consolidating 10+ disparate datasets (>1B rows), reducing dataset onboarding from 2 weeks → <1 hour with automated schema reconciliation

• Led 3-person team building S&P CapIQ data pipeline, mentored engineer to senior role: consolidated 50+ tables → 12 dbt models, optimized Snowflake clustering and right-sizing to cut analyst query complexity 95% and dev time 40%