Building scalable web apps, training local LLMs, and creating tools for fine-tuning, GPU inference, and multi-agent systems.

I'm a Senior Full Stack Engineer with 6+ years of commercial experience focused on building high-performance, production-ready systems - from SaaS platforms to real-time applications.

In the last years, I expanded my work into AI engineering, specializing in:

• 🧠 Local LLM inference (llama.cpp, GPU acceleration, CUDA optimization)
• 📚 Fine-tuning models with QLoRA, LoRA, and HF Transformers
• 🧩 Dataset engineering & enforcing structured reasoning formats
• 🛠 Building multi-agent architectures (AI-Station project)
• 🐳 Dockerized model deployment + FastAPI streaming backends
• 🔧 Custom tools for model merging, GGUF conversion, and orchestration

I work across the full stack - frontend, backend, DevOps, AI systems.

A toolkit for training, fine-tuning, merging, and converting LLMs locally.
Includes QLoRA pipelines, dataset builders, GGUF tools, Docker setups, and more.

➡️ Repo: https://github.com/sVarlog/LLMDataForge ⭐ Currently evolving into a reusable open-source framework.

A modular local multi-agent AI system with memory, workspaces, voice I/O, and GPU-backed inference.
Designed for productivity, coding assistance, automation, and experimentation.

➡️ Repo: Private (opens soon)

A full-stack recipe & grocery assistant platform:
Next.js frontend, NestJS backend, multi-language CMS, React Admin, Prisma, and Cloudflare R2 image pipeline.

➡️ Repo: Private MVP

• JavaScript, TypeScript, Python
• SQL (PostgreSQL), GraphQL
• Bash, CUDA basics

• React, Next.js, Vue, Nuxt
• TailwindCSS, PrimeReact
• Component libraries & UI optimization

• Node.js, NestJS, Express
• FastAPI (LLM backend)
• Prisma, PostgreSQL
• WebSockets, Redis, Bull Queues

• HuggingFace Transformers
• QLoRA, PEFT, LoRA adapters
• llama.cpp (CUDA, cuBLAS, GPU inference)
• Tokenization (BPE, SentencePiece)
• Dataset creation, prompt templates
• Streaming responses, KV cache optimization

• Docker + NVIDIA GPU containers
• CI/CD
• AWS, Vercel, Cloudflare R2

• LinkedIn: https://www.linkedin.com/in/mykhailo-nosov-574614183/
• GitHub: https://github.com/sVarlog
• Email: available upon request