Hi, I’m Kunj Shah — a Machine Learning Engineer focused on building, training, and optimizing Large Language Models. I work on LLM architecture, fine-tuning (SFT, DPO), and inference optimization at scale. I’ve built and deployed transformer models from scratch, including instruction-tuned and agentic systems. Currently studying Computer Science while leading AI initiatives and contributing to open-source research. Open to impactful collaborations in LLMs, agents, and applied AI systems.

• Website: https://kunjs.vercel.app
• LinkedIn: https://www.linkedin.com/in/kunjcr2
• Email: mailto:kunjcr2@gmail.com
• Hugging Face: https://huggingface.co/kunjcr2
• GitHub: https://github.com/kunjcr2

• Python, C++, JavaScript, Java, SQL, HTML/CSS

• PyTorch, TensorFlow, Keras, Scikit-learn
• Transformers, Large Language Models (LLMs)
• Fine-tuning (SFT, LoRA), Preference Optimization (DPO, PPO)
• NLP, Computer Vision, Reinforcement Learning
• Prompt Engineering, RAG Pipelines

• Hugging Face (Models, Hub), OpenAI API
• LangChain, LangFlow, n8n
• AI Agents, Tool Calling, Prompt Chaining
• MCP Servers, Vector Databases

• Docker, vLLM Serving
• Model Deployment, Inference Optimization
• GPU Optimization, Distributed Training
• Weights & Biases, Vertex AI

• Node.js, Express.js, Flask
• React.js, TailwindCSS, Streamlit

• MongoDB, MySQL
• Git, GitHub, Postman