Web platforms, mobile applications and AI agent systems — designed, shipped and operated on infrastructure we administer. Everything below is live or in build, and every claim opens in a new tab.
001 The work
Where the code is public the link goes to the repository. Where it isn't, the link goes to the running product.
Tanzanian employment law, with citations
Free question-and-answer over the Employment and Labour Relations Act, plus a compliance check that walks an employer through what the Acts actually require. Every answer cites the section it came from, so a reader can verify it against the law rather than trust the model.
KaziLaw is not a law firm and does not give legal advice. It is Tanzanian labour law, organised. kazilaw.com ↗A dinosaur chat app for children
Shipped on the Apple App Store and built contract-first: the OpenAPI schema generates the typed client, so the app and the API cannot quietly drift apart. Identity is derived on the device instead of collected, so a child never creates an account.
github.com/sugarshaneaz/DinoIG ↗Training, written for East Africa
Not translated into the region but written for it: Swahili first, bodyweight-first programming for people without a gym, and East African food in the nutrition side. Rotating trainer-led programs, set-by-set logging, and a licensed exercise library served from our own infrastructure instead of somebody else's quota.
Its programming engine is open source ↗Customer-facing AI with the guardrails written down
Each industry pack declares what the assistant must never invent and when it has to hand off to a person — the pharmacy pack refuses dosage, drug interactions and diagnosis outright. Blank onboarding fields are fed back into the prompt, so the model is told what it does not know instead of filling the gap itself.
github.com/sugarshaneaz/Sema-API ↗Deterministic program generation
The training-program generator extracted from Chuma and released on its own. Given a catalogue and a configuration it produces a full rotating program deterministically — same input, same program — with a test suite that asserts both full coverage where movement patterns exist and graceful degradation where they don't.
github.com/sugarshaneaz/routine-engine ↗002 Infrastructure
Two systems we built for ourselves and run in production. Neither is public — this page is the only place they are described.
The agent runtime our automation runs on. Long-lived agents with their own memory, scheduled work and tool access, sitting directly on our servers rather than behind someone else's platform — so the jobs keep their state and we keep the logs.
The agent that directs production work: media pipelines, batch generation, and the routine operational jobs that would otherwise be a person copying numbers between systems. It briefs the tools, checks the output and reports what it did.
003 Capabilities
Applications that do real work: search over your own documents, gated content, payments, reporting. Built on Next.js and Postgres, deployed and monitored by the same people who wrote them. See KaziLaw ↗
React Native apps taken all the way to the store — including the review process and the on-device bugs that never appear in a simulator. See Dino IG ↗
Retrieval over your own data, agents that run on a schedule, and explicit written rules for what the model must not say. See Sema-API ↗
004 Contact
Start with the problem rather than the spec — what should be true when it works, and who it has to work for. We will tell you honestly whether we are the right people to build it.
support@sylvesterlimited.tech