GenAI Engineer | Full Stack Developer | Building AI-Powered Applications | PyTorch, Transformers | Meta Hackathon Finalist
I am a B.Tech student specializing in Artificial Intelligence and Cloud Computing, with a strong interest in building real-world AI-powered applications. My work focuses on combining Generative AI, Machine Learning, and Full Stack Development to create scalable and efficient systems. I have hands-on experience with deep learning models, reinforcement learning, and modern web technologies, allowing me to develop end-to-end intelligent applications. I have built projects such as an AI-based portfolio optimization system using LSTM and multi-agent reinforcement learning (PPO, DDPG, A2C), a multi-tenant SaaS platform with secure authentication and billing integration, and an offline voice AI system using Whisper for real-time transcription. I enjoy working on performance optimization and system design, and have improved system efficiency and latency significantly across multiple projects. Achievements: • Finalist – Meta PyTorch OpenEnv Hackathon (Top ~2.6% among 31,000+ teams) • Winner – IBM Day Hackathon (AI/ML Track) I am currently looking for opportunities in Machine Learning, AI Engineering, and Full Stack Development where I can contribute to building impactful, real-world solutions.
United University
9 mos
Reinforcement Learning & Portfolio Optimization
Prerogative Group of Institutes (PGOI)
4 mos
MERN Stack
United University
09/2024 – 05/2025 · Prayagraj
Architected hybrid portfolio optimization combining LSTM temporal modeling with multi-agent reinforcement learning using PPO, DDPG, and A2C; engineered a state space with 7+ financial indicators; reduced prediction error by 15–25% and improved reward convergence by 25–35% over baselines; designed risk-aware rewards with Sharpe ratio, volatility penalties, and drawdown constraints; optimized training to 72–94 FPS, completing 50K-timestep cycles in approximately 31 minutes.
Prerogative Group of Institutes (PGOI)
09/2023 – 12/2023 · Remote
Built and shipped scalable MERN full-stack applications with modular backend architecture; designed secure RESTful APIs with JWT authentication and role-based access control; reduced backend response latency by 30–40% through query optimization and efficient data-access patterns; improved frontend load speed by approximately 25% through optimized rendering and state management.
Bachelor of Technology - BTech, Computer Science
Sep 2022 - Sep 2026
CGPA: 8.5. Relevant coursework: Data Structures & Algorithms, DBMS, Operating Systems, Computer Networks, Cloud Computing.
Class 12th - CBSE
Jul 2020 - Jul 2021
Class 10th - CBSE
May 2018 - May 2019
I built an AI-powered portfolio optimization system that leverages deep learning and reinforcement learning to make real-time investment decisions on market data. Used LSTM + PPO/DDPG/A2C for dynamic allocation. Integrated FinBERT for sentiment analysis. Designed custom OpenAI Gym trading environment. Achieved stable convergence over 50K+ timesteps. Real-time inference (<120 ms) using FastAPI + React.
Built a production-grade GenAI platform using Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG), agent tool-calling, and memory systems. Implemented FAISS-based vector search (1K+ chunks/user) for context-aware responses Developed multi-turn chat with real-time SSE streaming (<150ms latency) Integrated voice AI (Whisper + TTS) with ~50% latency optimization Designed full-stack system using FastAPI, React, PostgreSQL, Redis, and Docker
Built a production-grade GenAI platform with RAG, agent tool-calling, SSE streaming under 150ms latency, FAISS retrieval supporting 1K+ chunks per user, Redis and vector memory, voice AI using Whisper and TTS, and a FastAPI, React, PostgreSQL, and Docker stack. Improved response quality by 30%+ and optimized voice latency by 50%.
Built a Transformer and reinforcement learning system using PPO, DDPG, and A2C for portfolio optimization on real market data; designed a custom trading environment with Sharpe ratio and cost modeling; achieved stable rewards of approximately 68+ over 50K+ timesteps and enabled real-time inference under 120ms with FastAPI and React.
Built a multi-tenant SaaS architecture supporting 100+ users with CRM workflows, automation, and analytics; implemented RBAC authentication and Stripe billing; designed a scalable backend handling 1K+ API requests per day; improved workflow efficiency by 30%+ and reduced manual operations by 40%+.