Arshak T.

I build backend systems and AI integrations for platforms that need to actually work in production — not just a demo wired to an API key.

Most recently I was the Python backend developer on Mizou, an enterprise AI voice-assistant training platform. I built the Django REST API and database layer, integrated LangChain for prompt orchestration across multiple LLMs (OpenAI GPT-4o, Claude 3.5, Groq), added real-time audio/video with LiveKit and AI avatars via Tavus, and used Celery for background processing of transcripts, recordings, and evaluation scoring. The platform runs multi-tenant with role-based access, deployed on AWS Elastic Beanstalk with CI/CD, Stripe billing, and Sentry monitoring.

I've also led architecture and AI integration on Websage.AI (a RAG-based site assistant using OpenAI embeddings and a vector database) and Kodeops (an AI-powered project-management tool pulling from chat, task, and Git data via OpenAI). Alongside that, 12+ years of full-stack work in Python (Django, FastAPI) and PHP (Laravel), with React/Vue front ends, gives me a practical sense of how AI features actually plug into a real product — not just how to call an API.

Day to day I work inside modern AI-assisted dev environments — Cursor, Windsurf, and OpenAI Codex — and stay current as new models and tooling come out, so I'm not stuck integrating last year's stack.

What that means for you: I can take a vague "we want AI in our product" brief and turn it into a scoped backend architecture, or take an existing Django/FastAPI codebase and add LLM features without breaking what already works.