I build and ship production software across the stack, with a focus on backend systems, real-time AI, developer infrastructure, and Rust.

Currently working on production systems involving:

• ⚙️ Backend & distributed systems — service-oriented architectures, PostgreSQL, Redis, REST APIs, queues, and production infrastructure
• 🤖 AI infrastructure — real-time voice agents, LLM systems, MCP, and agent-native software
• 🦀 Rust & systems engineering — async systems, open-source infrastructure, Linux, and performance-sensitive software
• 🌐 Developer infrastructure — APIs, OpenAPI, WebMCP, debugging tools, and machine-readable software

I care about understanding systems end to end — from architecture and data flow to performance, deployment, and failure modes.

AI Agents
    │
    ├── MCP / WebMCP
    ├── Real-time Voice
    └── Tool & API Integration
             │
             ▼
      Backend Infrastructure
             │
    ├── Distributed Services
    ├── PostgreSQL / Redis
    ├── APIs / Queues
    └── Observability
             │
             ▼
       Systems Engineering
             │
    ├── Rust / Tokio
    ├── Linux / Wayland
    └── Performance

Yupcha AI Interviewer — Re-architected an AI video interviewer from a high-latency STT → LLM → TTS pipeline into real-time speech-to-speech using Python, Pipecat, AWS Bedrock Nova Sonic, LiveKit, and WebRTC. Designed the service architecture to support 100 concurrent live sessions, with asynchronous LLM scoring decoupled from live sessions. :chatgpt-content-reference{index="0"}

AHTML — Created a machine-readable HTML standard that generates OpenAPI 3.1, JSON-LD, llms.txt, and WebMCP from a single source. Published 9 npm packages with 25k+ combined downloads. :chatgpt-content-reference{index="1"}

Kasauti — Built an AI-powered assessment platform where an LLM authors questions, runs timed exams, auto-grades submissions, and provides per-topic analytics. Exposed the platform through MCP so agents can create exams, grade submissions, and generate targeted follow-up exams from weak areas. :chatgpt-content-reference{index="2"}

Fresco — Built and maintain an open-source Linux live wallpaper engine in Rust with a native GTK4 interface and optimized async playback pipeline. Reached 1,713 users across 115 countries and 5,695 downloads. :chatgpt-content-reference{index="3"}

WhatBroke — Built an AI debugging CLI and MCP server that packages crash context and uses Git history to identify suspect commits across multiple test runners. :chatgpt-content-reference{index="4"}

I actively contribute to production open-source projects, particularly in Rust.

• ReductStore — merged 8 PRs into a production Rust time-series database, including replication compression and observability work.
• Meta Pyrefly — fixed a false-positive Python type-checker diagnostic involving PEP 561 submodule handling.

Languages

Rust TypeScript JavaScript Go Python SQL

Backend & Systems

Tokio Node.js Hono FastAPI REST OpenAPI OAuth JWT

Data

PostgreSQL Redis MongoDB DynamoDB MariaDB

AI / Real-time

AWS Bedrock MCP AI Agents Pipecat LiveKit WebRTC

Infrastructure

AWS Docker Nginx Linux GitHub Actions CI/CD Systemd

Frontend

React Next.js TanStack Query Tailwind CSS

If my projects save you time, consider supporting their development. It funds maintenance, releases and new tools.

☕ Sponsor me on Ko-fi

Agent-native software · distributed systems · Rust infrastructure · real-time AI · developer tools · systems performance

If you're building something technically difficult, I'm interested.