I'm a student at NIT Rourkela deeply interested in Machine Learning, Generative AI, and software engineering. I enjoy building robust applications, fine-tuning language models, and continuously exploring new areas in tech.
I recently built Deal Hunter, a multi-agent AI system that scans RSS electronics feeds, extracts structured product opportunities, and ranks deals by estimating fair price and discount potential. This project combines:
- Agentic Orchestration: Scanner, planner, and ensemble agents that coordinate extraction, prioritization, and notification workflows.
- Hybrid Price Intelligence: Frontier LLM inference with retrieval plus a Modal-hosted specialist model, merged through a configurable ensemble.
- Production Components: ChromaDB memory, Pydantic-typed models, config-driven services, and a Gradio dashboard for live monitoring.
I recently built Product-Pricer, an end-to-end MLOps pipeline that predicts Amazon product prices from raw metadata. This project benchmarks AI's ability to "understand" value by comparing:
- Specialized Fine-Tuning: Custom trained adapters for open-source (Llama 3.1) and commercial (GPT-4o Mini) models.
- Frontier LLMs: Zero-shot inference using Claude, Gemini, and OpenAI models.
- Traditional ML: Baselines using Random Forest, SVR, and Linear Regression.
- Programming: Python (OOP, DSA)
- AI & Machine Learning: PyTorch, Hugging Face Transformers, Scikit-learn, PEFT (QLoRA), Langchain
- MLOps & Tools: Docker, Hydra, Pytest, Loguru, Git, Modal, ChromaDB, Gradio
- Systems: Linux, Virtualization
- Architecting multi-agent systems, ensemble models, and autonomous planning workflows for complex decision-making tasks.
- Refining problem-solving skills through advanced Data Structures and Algorithms in Python.
📫 Connect with me: [visheshpachouri@gmail.com]