Yatharth A.

Credit Risk Scoring & Probability of Default Model

I built a Logistic Regression/XGBoost-based PD model on 2.2 million+ of historical loan data, achieving AUC of 0.75. Translated probability of default into 4-tier Basel II/III credit decision bands, enabling risk segmentation and analysis.

Conversational RAG Engine for Interactive Document Analysis

Developed an interactive LLM assistant (similar to Gemini), featuring a conversational memory buffer. Implemented a document ingestion system, achieving 94% Retrieval Precision by optimizing semantic chunking strategies.

AI-Driven Software Engineering Automation Framework

Designed AI-enabled framework using Claude Code and Codex to automate software development and QA workflows. Defined DevSecOps workflow integrating Git PRs, CI/CD, SAST, secret scanning, dependency checks & approvals.