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