Sumit

Chatterjee

Pioneered 16-bit HDR EXR generation from AI models using exposure bracketing and custom LoRA training. Developed ComfyUI nodes that preserve full dynamic range.

Implemented Foundry Nuke's professional compositing nodes for ComfyUI, bridging traditional VFX workflows with AI generation.

Neural network approach to expand 8-bit images to higher bit depths using U-Net architecture for professional VFX use.

Expert in training custom LoRA and full fine-tuning of diffusion models using KohyaSS, AI-Toolkit, and Diffusion Pipe for specialized image generation.

Deploying and optimizing local LLM inference using vLLM, Ollama, and LM Studio for high-performance, privacy-focused AI applications.

Maximizing productivity with AI code editors including Cursor, GitHub Copilot, Cline, Kilo Code, Gemini CLI, and Claude Code for accelerated development.

Containerizing AI/ML pipelines, development environments, and production deployments with Docker for reproducible and scalable workflows.

Building autonomous AI agents using LangChain, LangGraph, CrewAI, and Microsoft AutoGen for complex multi-step reasoning and task automation.

Creating sophisticated automation workflows connecting AI services, APIs, and business processes using the n8n platform.

Creating and customizing Model Context Protocol servers to extend AI assistant capabilities with custom tools and integrations.

Fine-tuning large language models using SFT, DPO, RLHF, and QLoRA techniques for domain-specific applications and improved performance.

Efficient handling, streaming, and processing of large-scale datasets from HuggingFace Hub for training and evaluation workflows.