I’m Turuu Byambaa, a software builder and open-source maintainer in Ulaanbaatar. I work on products before the idea has become a clean ticket: when the domain is unfamiliar, the useful boundary is still unclear, or the technical model no longer matches reality.
I learn the domain, find the real boundary, make the product and architecture coherent, help the team grow around the work, and stay through production and operations.
Independent products · AI infrastructure · Complete systems · Open source
Current public work
Open Apps began in 2017 as a list of complete open-source applications—the kind of codebase I used to learn how real products fit together. In 2026, I rebuilt its maintenance model around structured records, automated validation, generated output, and a clearer contribution path.
The repository already had an audience. The current work preserves that history instead of claiming it as a new result, then addresses the maintenance problems the original list could not solve.
Public snapshot · 2026-07-29
Open case studySelected evidence
Each story shows a different difficult boundary: production AI, enterprise delivery, a venture under real demand, and developer products maintained in public.
Learn the unfamiliar stack, then make it usable by the whole team.
Worked on customer-facing and internal banking systems while turning new frontend and cloud approaches into shared delivery practice.
Open-source lineage
The point is not a repository count. Public work made incomplete understanding visible, then added the harder obligation: keep the code useful after its first release and understandable beyond its original author.
Early applications made data flow, errors, product structure, and deployment visible end to end.
Intelligo, its CLI, Neuro.js, examples, and bilingual documentation turned private learning into a public interface.
Open Apps carries the product evidence. Grove carries only the reusable maintenance mechanics the product earned.
Fleetbase required mapping boundaries across ten repositories and getting 40 pull requests accepted upstream.
How I work
Find the actual boundary of the problem—product, data, operations, and failure modes included—before choosing the implementation boundary.
Mazaal platform work · Tixy product and company boundary
Learn unfamiliar work against a real delivery, turn the reasoning into examples and review standards, and make the next decision less dependent on one person.
Khan Bank team practice · Intelligo documentation
Repeated pain inside a working product is evidence for an abstraction. A framework still has to earn independent use.
Open Apps → Grove
Contact
The most relevant conversations are about an ownerless system, a product stuck between prototype and production, an architecture crossing product, data, AI, and operations, or a technical team that needs a clearer working standard.
Write with the constraint, what has already been tried, and what a successful change would make possible.