Developed a full-stack AI system that transforms unstructured web data into a searchable knowledge base. Integrated automated scraping, chunking, embedding generation, and retrieval-augmented generation (RAG) to deliver accurate, context-aware answers via a FastAPI API.

Developed a scalable food ordering application for BeFoodly with a React frontend and Java Spring Boot backend. Implemented RESTful APIs for order management, optimized data flow between services, and ensured a responsive and reliable user experience across the platform.

AI-powered systems that enable natural, human-like conversations through spoken language by combining speech recognition, large language models, and speech synthesis in a low-latency streaming pipeline. They can listen, understand, and respond almost instantly, allowing users to interact hands-free while accessing knowledge bases, calling APIs, and performing tasks like booking services or retrieving personalized data. By integrating technologies such as Whisper for speech-to-text and advanced text-to-speech engines like ElevenLabs, these agents create seamless, conversational experiences across applications like customer support, virtual assistants, and voice-enabled automation systems.

Built a Figma/Excalidraw-inspired diagram editor from scratch — canvas-based shape rendering, drag-and-drop editing, and a Zustand-driven state layer on the frontend, backed by a Python service for persistence.

A multi-agent FastAPI service that evaluates learner artifacts — code, essays, briefs — against an inferred rubric. Runs a rubric-builder, critic, gap-detector, and synthesiser agent pipeline powered by LangChain and Google Gemini to produce scored, evidence-backed verdicts.

A low-level system design implementation for a rental/booking service in Java — focused on clean OOP architecture, extensibility, and the SOLID principles interviewers look for in machine-coding rounds.