I read one or two repos, interview a handful of engineers, and hand back a written score of how agent-ready the codebase is, a prioritised fix list, and a live demo of an agent doing a real ticket in your repo, before and after a day of harnessing.
Alon Ohad
Engineering Manager who ships
I have spent my career building software, and close to a decade leading the teams that build it. I still write code every day. Right now that means running coding agents inside production monorepos and getting the repository into a state where what they produce is something you would actually merge.
How I work with agents
Most engineering orgs bought Copilot or Cursor seats and watched their seniors reject the agent-written pull requests. The seats are fine. The repository is the problem.
Ways to work together
I build the rule sets, the CI contract enforcement, the MCP servers, and a per-team workflow, then train the team in-repo on real tickets rather than slides.
I keep the rules and MCP servers current as models and your stack change, and review agent-generated pull requests.
Scope and pricing depend on the codebase. Tell me about yours. See a real repo scored before/after, with the actual harness changes and one ticket run cold on both trees.
Track record
- ironSourceCore servers and the department's BI system, at scale.
- Sensi.AIFull-stack team lead for the platform behind audio-based care insights.
- TastewiseLed backend and frontend for the food-intelligence platform and shipped its AI features into the product.
- XpendAI-driven engineering for a multi-brand B2C platform: LLM features, agentic workflows, runtime brand configuration, growth systems, one codebase.
Stack
Also built
Away from the screen
I have guided sea kayaking since 2000 and led expeditions in places like Tasmania and Svalbard. The rest of this site is about that.