I'm Franklin Ssozi. I work with engineering teams that have stopped delivering. I fix the architecture, the team structure and the delivery process, and I install open source AI tools that keep your team shipping after I leave.
The assessment finds out which one is stopping your team, and then I fix it with your team.
“We've been building but have nothing to show for it.”
The team has the wrong shape for the work, and no one decides what ships first. I rebuild the team around the first release and give it a delivery process that everyone follows.
“Everything is tightly coupled. We can't validate units in isolation.”
Every component shares state with every other component, so a small change means retesting the whole platform and teams wait on each other. I split the platform into parts with hard boundaries and contracts, so each team tests and ships its part alone. AI tools also write better code when they can see where one part ends and the next begins.
Work passes between people by hand. I set up a process where people agree on the spec before work starts, AI does the drafting, and people review the result against written rules before anything ships.
Six engineers spent a year building a dealer analytics platform and shipped nothing. Each dealer had its own Python ETL code, so every new dealer and every change needed a developer.
I restructured the team, replaced the per-dealer code with one shared pipeline, and moved the team onto an AI native delivery process. Analysts now read and change the data models, and a metrics registry defines every number the product shows.
The platform shipped 4.5 months after I started, and dealers use it every day. One engineer and the business analysts ship new features in days. The metrics registry also feeds an AI assistant inside the product that answers dealer questions.
Over 60 engineers worked on one platform where every component read shared configuration. A change took hours to write and a week to verify, because no team could test its part without testing everything.
I designed an architecture that splits the platform into self-contained capabilities with contracts between them. Rarely changed configuration moved into code, and every API now has a clear schema that AI tools can read.
Each capability tests on its own, so teams ship without waiting on each other, and a scoped change verifies in days. New client requests fit an existing extension point instead of forcing a redesign.
Five freelance teachers took 6 months to write each course, and the work needed several rounds of revision. The company had no AI expertise, did not trust AI-written content, and ran an app whose APIs were not built for AI.
I designed a content pipeline where AI drafts the course, a subject analyst and a designer own it, and 2 to 3 teachers review it before release. Every claim carries its source citation, and automated checks confirm each source says what the content claims. I also designed the architecture for the new platform and app, and the content still runs on the old app for students with older devices.
The company produces a full course in about a week, and has done it across several syllabi.
Every engagement installs the same toolkit in your repos. It's public on GitHub, so your engineers can read it before we talk, and it stays with your team after I leave. You pay for my time to fit it to your system and teach your team to run it.
An AI native delivery process that runs inside your repos.
Produces the documents that architecture decisions and executive reviews depend on, all to one writing standard.
I read the code, map how work moves through the team, and interview the people doing it. You get a written plan that says what blocks delivery, what to change, in what order, and what it costs. The plan is yours, and you can run it without me.
I deliver one change with your team, such as an architecture split or a new data pipeline, and install the toolkit along the way. The project ends when the change ships and your team runs the toolkit without me.
I act as your architect or head of engineering, part time, for as long as you need me. We review the arrangement every quarter.
I've spent years building loan origination and dealer finance systems, where a bad record costs real money.
I run Inflection Agents as a one-person practice, so the person who assesses your team is the person who does the work.
Tell me how many engineers you have, what you're building, and what has stopped shipping.
kiwanuka@