Neha Shastri

About

Backend/ML engineer who built a job-search automation SaaS using a modular Selenium ETL pipeline, rigorous testing/observability, and a cost-optimized two-pass LLM ranking approach. Has led high-integrity data extraction from messy multi-city PDF records (95% integrity) and managed modular production rollouts for a 20+ engineer team, with a strong security focus (deny-by-default, row-level access control) in an AI-assisted grading platform.

Key Strengths

• Built a scalable job-search automation SaaS with modular Selenium-based ETL pipeline
• Strong testing discipline (unit + integration) and observability via logging
• Cost/performance optimization via two-pass ranking (cheap first pass, stronger model rerank)
• Iterative LLM quality improvement through monitoring and prompt/token optimization
• Led data extraction/refactor for messy PDF records across five major U.S. cities; achieved 95% data integrity
• Managed production GitHub repo and rolled out modular updates to a 20+ engineer team
• Designed robust ETL safeguards: automated schema-change validation, entity deduplication, and retry/fallback for failed API calls
• Security-first approach: deny-by-default permissions, FastAPI request verification, and row-level access control for sensitive student data
• Identified imbalanced-data failure mode and improved ML performance by ~80% using a two-step classification-then-regression approach

Experience

Education

Awards

• Finalist (Top 5) - MinneMUDAC Data Challenge

Skills

Languages