Samuel Birhanu
software engineer — agentic AI, Addis Ababa, Ethiopia
Agent Barn
I'm a core contributor to Agent Barn, a control plane for running fleets of AI agents on your own Kubernetes infrastructure instead of somebody else's cloud.
My part of it: the Python APIs that define agents and what they can touch, the Helm charts that deploy the whole stack, LiteLLM model routing with per-agent cost attribution (so you can see exactly what each agent spent), and the skills that hook agents into Slack, Teams, and Telegram.
The reason I care about this setup: once agents touch real customer data, "just trust the hosted platform" stops being an easy answer. Agent Barn keeps the agents, the tools, and the data on infrastructure you control.
Klikt
I worked on Klikt as a software engineer. It's an AI research analyst for B2B sales teams: it runs a deep-research pipeline over each prospect, scores them against your ICP, and writes out why it scored them the way it did. A bare score never helped anyone; the reasoning is the part you can check.
Results export to HubSpot, Pipedrive, or a plain CSV.