I'm Sekou. Twenty years turning business strategy into distributed systems, from multi-region networks and disaster recovery to agent infrastructure and local model serving. This is where I think out loud about that work: the trade-offs, the failures, and what actually held up in production.
Sekou M. Doumbouya, infrastructure/cloud architecture leader. Not the basketball player.
01. Writing
Essays on systems design, infrastructure leadership, and AI workloads.
I ran official Qwen3.8 27B BF16 and FP8 side by side, then changed one serving knob at a time. The fastest result was not the whole story.
Read article →The best long-context profile I measured passed every synthetic gate and crashed twice under real traffic. Here is why promotion has to be a guarded transaction, not a config edit.
Why I added a second RTX PRO 6000, upgraded my CPU, and then watched DeepSeek 0731 land on the exact card I just bought.
A 41-task graph ran for 41 hours, produced 115 reviews, and merged nothing. Then I pivoted mid-flight to milestones, and the same session shipped five PRs in half the time.
02. Proof of Impact
A few case studies, when you want the receipts behind the writing.
Coding harnesses can point at a local model, but most routers decide by static model names, costs, or regexes. That is not enough for agent work. A model can return clean JSON that is still semantically wrong, and a bad plan can poison an entire run without throwing an error.
Read case study →Multiple AI coding agents and the humans supervising them increasingly work the same plan at the same time, but the tools they coordinate through (GitHub Issues, markdown conventions, chat history) store free-form text, claim work by label or a chat message, and enforce nothing when two agents grab the same task. Worse, when an agent reports a task complete, nothing verifies the claim. Status outranks evidence, which is exactly backwards for autonomous or weakly-supervised runners.
Read case study →A 100% TypeScript engine on Bun that runs Claude-powered agent tasks with conversation-level checkpointing, pluggable sandboxes, and a 27-tool MCP server.
Read case study →03. Community
Thought leadership, community, and open source.
Interested in AI infrastructure, platform engineering, or just want to connect? I'm always happy to talk shop.
Get in Touch →
