Projects

Opsready AI Agent

Built a functional MVP for OpsReady, an AI-powered work order management system designed to replace traditional search with a natural language interface. Developed a tool-calling AI agent using Claude API that enables conversational execution of workflows and real-time data queries. Implemented backend logic for 14+ custom tools supporting API-based interactions, allowing the agent to perform structured actions across operational systems. Designed a modular architecture for scalable tool integration and extensible workflow automation.

Tech Stack: Python (FastAPI), React, Claude API (Anthropic), Tool Calling / Function Calling, REST APIs

LinksV8

Developing a full-stack social matching platform designed to connect users based on location and shared interests. Building a custom backend hosted on Azure to replace BaaS dependencies and enable full control over system architecture. Implementing a PostgreSQL database with PostGIS for geo-spatial queries, enabling proximity-based event discovery and user matching. Developing JWT-based authentication and laying the foundation for a scalable recommendation system and real-time communication layer.

Tech Stack: Flutter (Dart), C# (ASP.NET), PostgreSQL, JWT Authentication

Cinemax

Built a movie recommendation platform using a hybrid recommendation engine combining collaborative filtering (SVD) and content-based ranking to generate personalized film suggestions. Designed a full-stack system using Python backend services and React frontend, with PostgreSQL for persistent user data. Actively used by a small group of users for exploring and curating personalized movie recommendations, with a focus on improving discovery of niche and lesser-known films.

Tech Stack: Python (Flask, NumPy, Pandas, SciPy), React, PostgreSQL, JWT Authentication

FeelScape

Built an AI-powered music generation system that creates ambient LoFi soundtracks based on visual input from user-uploaded images. Used computer vision techniques to extract color and composition features to guide audio generation. Developed a full-stack system integrating FastAPI backend services with React frontend, enabling automated generation of mood-matched audio loops for productivity and focus use cases.

Tech Stack: Python (FastAPI, OpenCV, NumPy, Scikit-learn), React



Contributions

SecureAI - Open Source Contribution

LLM Security Evaluation Output
                Model: huggingface/gpt2
                Containerized Execution: Docker
                Probe Categories Run: 3
                Detected Vulnerabilities: 0
                
                Evaluated Failure Modes:
                - Data Leakage
                - Prompt Leakage
                - Token Extraction

Worked with a Docker-based LLM security evaluation framework to execute structured probe and detector suites across multiple vulnerability categories, including prompt leakage, data leakage, and jailbreak testing. Analyzed model behavior and evaluation outputs to understand LLM failure modes and safety limitations.