Embedded Files
🔍 LLM-Powered Testing Smell Detection Tool
🔍 LLM-Powered Testing Smell Detection Tool
Duration: Jan 2025 – Apr 2025
Tech Stack: Python, OpenAI GPT-4, AST Parsing, GitHub Actions, CI/CD
Description:
Engineered an automation tool that uses Large Language Models (LLMs) to analyze test files and detect test code smells like Assertion Roulette, Eager Test, and General Fixture patterns. Integrated with GitHub CI pipelines for on-push analysis and auto-commenting on PRs.
✅ Increased test coverage by 30%
✅ Reduced execution cycle time by 20%
✅ Enhanced explainability with inline suggestions powered by GPT
🛡️ AI-Driven Edge Risk Detector
🛡️ AI-Driven Edge Risk Detector
Duration: Aug 2024 – Dec 2024
Tech Stack: Python, TensorFlow, OpenCV, Raspberry Pi, Edge ML
Description:
Developed a real-time risk detection system deployed on edge devices to identify falls and critical events in constrained environments like hospitals and elderly care.
✅ Achieved 95%+ accuracy in event classification
✅ Reduced inference latency for live edge video feeds
✅ Designed an optimized model pipeline for low-power hardware (Raspberry Pi)
💸 Finance Buddy – AI Web App (Hack Dearborn 3)
💸 Finance Buddy – AI Web App (Hack Dearborn 3)
Role: Core Developer (collab with Akash Mahajan)
Duration: ~6 months ago
Tech Stack: Azure-hosted GPT‑4o, Splitwise API, Flask/React, Cloud Deployment
Description:
Built a gamified AI-powered financial health platform for Hack Dearborn 3 called Finance Buddy, integrating Splitwise data to generate personalized money-saving challenges via GPT-4. Deployed on Azure Cloud, the app delivers real-time insights and practical tips, combining financial planning with engaging interaction.
🛠️ Secure Insurance (AI-driven prototype)
🛠️ Secure Insurance (AI-driven prototype)
Role: Concept Developer | Independent
Duration: Late 2023
Tech Stack: Python, scikit-learn, Flask, SQLite
Description:
Built an AI-based insurance claim risk classifier prototype as part of an experimental research initiative. The system used tabular historical data to predict fraudulent or risky claims, offering explainable outputs using SHAP values to support transparent decision-making.
Page updated
Report abuse