Skip to work
KAYES
Kayes // AI Builder

Kayes Fardows — AI/LLM Application Developer. I buildSystemsthat Think.

I build AI and LLM-powered applications that feel like magic. Assistants that remember, agents that act, and products that survive real users and real load.

Read CVDownload CV
Oracle Presenceawake

The Oracle awakens. Kayes' work, craft, and stack are mine to reveal. What do you seek?

speak to the presence - it remembers

伝

Forged, not assembled

/LLM INTEGRATION/RAG PIPELINES/AGENT ORCHESTRATION/VECTOR SEARCH/TOOL CALLING/STREAMING APIS/EVAL HARNESSES/PROMPT ENGINEERING/FASTAPI/TYPESCRIPT

AI / LLM Projects

A selection of AI and LLM-powered applications I've built. Each one goes beyond a notebook prototype — real interfaces, real users, real constraints.

01

Multi-interface AI assistant running a unified brain across Web, TUI, CLI, and Telegram.

Agens

Agens is a free, self-hosted AI assistant that connects a centralized intelligent brain to a modern Web dashboard, Terminal UI, CLI, and a Telegram bot. It features native web search, safe local shell execution, cryptographic key storage, and automatic failover handling to seamlessly recover from rate limits.

  • Python
  • Svelte 5
  • Vite
  • FastAPI
  • SQLAlchemy
02

A local-first, privacy-focused RAG system with interactive citation grounding and exact-region PDF highlighting.

Ambrew

Ambrew is a local-first, privacy-focused research assistant that lets users chat with their documents while ensuring complete citation traceability.

  • FastAPI
  • Python
  • React
  • TypeScript
  • Tailwind CSS
03

AI-Powered Learning & Productivity Platform

LearnSync

A full-stack AI learning and productivity platform combining Next.

  • Next.js
  • TypeScript
  • FastAPI
  • Python
  • LangGraph
All Projects

The stack

What I reach for when the model is the easy part. Tools chosen for the boring reasons: they hold up, they observe well, and they fail loudly instead of silently.

Applied AI

  • Multi-agent orchestration
  • Retrieval-augmented generation
  • Tool calling & function APIs
  • Local + hosted model serving

Backend

  • FastAPI / Python
  • Node.js / TypeScript
  • Streaming & WebSockets
  • Auth (JWT, sessions)

Data

  • PostgreSQL
  • Redis
  • Vector stores (Qdrant, Chroma)
  • Schema-first extraction

Workflow

  • Docker
  • Linux (Arch)
  • Neovim
  • CI & evals

Got a model that
needs a system?

I take on a small number of builds where the AI has to actually work in production. If that sounds like your problem, start a thread.

kayes@runtime: ~/contact

$ open channel

Email Kayes