š Full-stack engineer, Python and TypeScript/JavaScript, 20+ years building production software
š¤ Specialized in applied AI: LLM agents, RAG systems, and voice assistants that run in production, not just in a demo
Tired of AI demos that wow everyone in a screen-share, then fall apart the first time a real user touches them? Or "cutting-edge" prototypes you end up babysitting that never quite reach production?
That's the gap I close. I treat AI features as real software, not a wrapped API call: eval sets before anything ships, logging you can actually debug, model routing to keep the bill sane, and a human in the loop wherever a wrong answer is expensive. Whether you need an LLM agent, a RAG pipeline over your own data, a voice assistant, or the full web app around it, I build the version that still works after the demo is over.
šØāš» Core skills and technologies:
- AI and LLM: OpenAI/GPT, Claude, Gemini; RAG (hybrid search, custom embeddings); agents, tool-calling, structured outputs; eval harnesses; prompt caching; fine-tuning
- Voice: Deepgram, Twilio, real-time STT ā LLM ā TTS pipelines
- Backend: Python, Django, FastAPI, Node.js (Express, NestJS), Celery, PostgreSQL, Redis
- Frontend: TypeScript, React/Next.js, Svelte/SvelteKit, Vue/Nuxt
- Cloud and infra: IaC, AWS, Cloudflare, Docker, CI/CD
- Generative media: image and video generation, both hosted (Nano Banana, GPT Image 2, Seedance 2) and self-hosted open-source (Qwen-Image, Wan) in ComfyUI with custom LoRA training
Send me a few lines about what you're building and where it's stuck, and I'll tell you straight what it'll take. š©




azamat.ai is my own AI engineering practice. Most of the time that means me; sometimes one or two engineers I trust. I build productio...
azamat.ai is my own AI engineering practice. Most of the time that means me; sometimes one or two engineers I trust. I build production LLM, voice, and RAG systems for founders and small teams in the US, Europe, and Central Asia, and the part I care about most is the one people skip: keeping them running once the demo is over.
Biometric is a technological startup from Kazakhstan, which provides biometric identification and facial recognition solutions. Hired...
Biometric is a technological startup from Kazakhstan, which provides biometric identification and facial recognition solutions. Hired and lead a team of 10 developers. Managed development, UX, analytics and feature specs. Designed a microservice oriented architecture based on a docker cluster spanning over 20 hosts. Continuously trained CNN for facial recognition task based on frequently updated dataset.