Neha Chaudhari

At Thingbits I took an agentic support assistant from prototype to production. It handles roughly 1,000 queries a week and resolves 60% of manual customer support end to end, answering from live order and catalog data through five tools instead of the model's memory. I also cut p95 latency 40% and shipped the REST APIs and vector search underneath it.

Master's Degree

Clark University

August 2024 - December 2025

Field: Computer Science

Coursework: LLM-Based AI Assistants, Distributed Systems, Algorithms. Award: 2nd place, Clark Tech Innovation Challenge.

Advanced

Intermediate

Mitig8it

January 2026 - Present

• Built Mitig8it from scratch, a GitHub-native platform that flags likely exploitable vulnerabilities in pull requests before merge, with inline fixes posted in under a second. • Architected the full system: FastAPI and Node backend, React and Vite frontend, GitHub App integration, and the security analysis pipeline. • Raised precision and cut false positives with a 3-tier pipeline: regex detection across 35+ CWEs, OpenGrep and Semgrep AST analysis across 8 languages, then LLM triage. • Engineered a repository profiling engine (framework, auth, validation, database patterns) for context-aware analysis, and mapped findings across CWE, OWASP, ATT&CK, and CAPEC. • Measured triage quality with an evaluation harness over a labeled golden-PR dataset (precision, recall, false-positive rate) plus OpenTelemetry tracing to catch model regressions before deploy.