← Home

Projects

AI-Powered Resume Evaluator & Hiring Decision Engine

Feb 2026 – May 2026 · GitHub

Tech Stack: FastAPI · Python · React.js · Gemini / OpenAI API

Built an evidence-based candidate evaluation system that parses resumes and job descriptions, scores candidates using a weighted 100-point rubric, and generates hiring recommendations with SWOT analysis.

Design decisions:

  • Designed a validation layer to verify evidence quotes, prevent keyword stuffing, and enforce score bounds — deterministically recomputes final scores for reliability and consistency over LLM hallucination
  • Developed adversarial testing to benchmark scoring quality and surface keyword-gaming candidate profiles
  • Deployed a full-stack MVP supporting PDF, DOCX, and TXT uploads with recruiter-facing dashboards

The central challenge was building a system that is simultaneously rigorous (evidence-grounded, reproducible) and practical (works on real-world, inconsistently formatted documents). The validation architecture was designed to make the system robust against the specific failure modes of LLM-based scoring: confident wrongness and score drift.


Capital Allocation Strategies across Domestic and Global Markets

Aug 2024 – Dec 2024

Designed a quantitative decision framework for capital allocation under varying risk appetites — documented in the Finance section.