About
Yad Konrad - ML systems, coding agents, program synthesis, security.
About
Hacking on new things for fun and learning. I make machines write code. Then ship it.
2x CTO, 2x Head of ML. 10+ years building ML systems, data infrastructure, and AI products. I’ve built data platforms from scratch and shipped production systems across startups and foundations. Currently I build agent-driven systems, data pipelines, and whatever else needs to exist.
What I Do
AI Coding Agents — LLM workflows, multi-agent orchestration, RAG systems. I’ve been trying to make computers write code since 2017. Now I build agentic platforms for enterprise clients and write about it at programsynthesis.pub.
Data Infrastructure — Pipelines, warehouses, knowledge graphs. ETL systems, migrations, making messy data usable. The hard part is never scale, it’s understanding what the data means before you touch it.
ML in Production — Training models is the easy part. Shipping them is the job. Reinforcement learning for physical systems, recommendation engines, graph neural networks.
Full Stack — Backend, frontend, infra, deployment. When needed, I build the whole thing.
Tinkering
What I’m currently building:
- CodingAgents.Dev — Link aggregator and community for coding agents, built with a coding agent
- AgenticEngineer.ING — Interactive lessons for becoming an agent-driven engineer
- Erdős Navigator — Agent navigator that turns Claude Code into a theorem proving assistant
- yad.codes — This blog, a digital garden
Background
KayOS (2024-2026) — CTO & Co-Founder. Built 13 production AI platforms across legal tech, marketing analytics, and BI. Agent infrastructure including knowledge graphs, AI sandboxes, ETL frameworks, and a full AI coding IDE with ACP integration. 21k+ commits across 65 repos.
NEAR Atlas @ NEAR Foundation (2022-2024) — Senior Data Engineer & ML. Data infrastructure for blockchain analytics. 770M+ transactions processed.
Be On Deck (2019, 2020-2022) — Senior ML/Data Engineer. ETL, data infrastructure, and matchmaking algorithms for founders.
Roby (2018-2019) — Head of ML. Reinforcement learning for HVAC optimization.
South Park Commons (2019-2020) — Indie Researcher. Program synthesis, Mora Search Engine.
Draper TV (2013-2015) — Technical Lead. Video infrastructure for online education.
Interests
Self-Healing and Improving Agent Systems — LLM long-horizon tasks, vision and world models, security research, sandboxes, VMs. How do you build systems that fix themselves and get better over time.
Program Synthesis and Coding Agents — The thing I keep coming back to. Making machines find programs from specifications. I write about this at programsynthesis.pub.
RL and Neuroscience — Curated 432 papers in Study-Reinforcement-Learning. The intersection of how brains learn and how we train agents.
Comedy and Creative Writing — Built Library of Comedy, the internet’s largest standup comedy database. I’m also a comedian. Yes, really.
Autodidacticism — I learn things by building things that are slightly out of scope and figuring out why they broke.
Open Source
Early contributor to Keras, TensorFlow, and Theano (2015-2017).
- One-Hundred-Layers-Tiramisu — Semantic segmentation
- Nvidia-Autopilot-Keras — End-to-end autonomous driving
- Study-Reinforcement-Learning — Curated RL and alignment papers
- Hopper — AI coding assistant, built in 2017
Research
- programsynthesis.pub — Making computers write code
- Study-Reinforcement-Learning — 432 papers on RL, RLHF, LLM alignment
Get In Touch
- GitHub: 0bserver07
- Twitter: @yadkonrad
For work inquiries: yadkonrad.com/workwithme