Michael Cengkuru · Open Data Specialist

I work where records meet decisions.

A system can publish hundreds of fields and still leave the next question unanswered. I build publishing systems, trace gaps in data, and connect sources so people can examine a claim and decide what follows.

Built: Mozambique · Built: Kaduna · Substantial development support: Zambia · Developer contribution; later review and improvement support: Uganda.

Built: Mozambique Built: Kaduna

Selected work

What I built, and where I helped.

A public portal, a missing field, and a disputed decision require different work. These examples separate what I built from the development, review, and advice I contributed to other teams.

Open infrastructure data could not be seen in one place

The problem: Twelve governments and agencies publish infrastructure projects in the OC4IDS standard, each on its own portal. Nobody could see them together, or tell which published locations were real sites and which were defaults, swapped coordinates or bare place names.
My contribution: I built the Open Infrastructure Atlas, an independent prototype that reads every publisher's feed, tests each location claim and draws only what the evidence supports.
What you can check: open-infra-atlas.pages.dev and the project page: 42,361 records, of which 275 place a project exactly as published.
Verified state: Publicly reachable; data snapshot of 24 September 2026 (checked 30 September 2026). A prototype, not an official CoST or publisher product, and not a claim of adoption or measured impact.
Source and date: The atlas's own published data file, snapshot of 24 September 2026.

The road fund's projects were not publicly readable

The problem: Fundo de Estradas holds Mozambique's road and bridge programme in internal systems. A citizen, journalist or oversight body had no public route to follow a project from identification to delivery.
My contribution: I built the public infrastructure disclosure portal.
What you can check: dadosinfra.co.mz, a bilingual project-information portal with project records, an open-data page and a publishing interface.
Verified state: The portal is publicly reachable and states in its own notice that it was developed by Fundo de Estradas and validated as an OC4IDS publisher in 2026 (checked 21 September 2026). Fundo de Estradas is the institutional owner. This is a record of what was built, not a claim of sole authorship, adoption or measured impact.
Source and date: Fundo de Estradas portal, checked 21 September 2026; personal role from the direct supplied brief recorded in the internal claims ledger for this page.

A state authority needed a public surface for its projects

The problem: KADPPA, the Kaduna State Public Procurement Authority, needed a public place to publish infrastructure project information that its own officers and the public could both read.
My contribution: I built the infrastructure project-information portal.
What you can check: ipdata.kdsg.gov.ng, the authority's public project listing.
Verified state: The address responds and serves the public portal route (checked 21 September 2026). Its listing is rendered in the browser, so this check establishes the published artifact, not the currency of the records inside it or its current maintenance.
Source and date: KADPPA portal, checked 21 September 2026; personal role from the direct supplied brief recorded in the internal claims ledger for this page.

In both cases the portal is evidence that an artifact exists and is public. My contribution to it is evidenced separately, by the brief recorded in the claims ledger beside this page, because a live portal on its own cannot show who built it.

Development support

For NCC and CoST Zambia, I provided substantial support to the development of the Infrastructure Publication Portal.

For CoST Sekondi-Takoradi, Ghana, I contributed development input and feedback to the information platform.

Review and improvement

I contributed as a developer to Uganda's Government Procurement Portal, and later supported review and improvement work on it, including the June 2026 relaunch.

In review work, I examine what the existing system shows, where the record breaks, and which change would help its users.

Advisory and data quality

A published record can leave gaps in the questions it was meant to answer. I trace those gaps before recommending another feature or dataset.

For Malawi's IPPI, I provided technical advisory and data-quality support, including indicator design, publication diagnostics and integration assessment. This was advisory and diagnostic work, not building or maintaining the platform.

Connecting sources

For an anonymised energy-sector dam-monitoring initiative, I developed an early-warning workflow linking satellite displacement, rainfall, field inputs, alerts, and reports.

The work brings different kinds of evidence into one workflow, while keeping their limits visible. This describes development work; operational acceptance, ownership and use are unverified and are not claimed here.

These are my contributions within institutional and team efforts. Words are used deliberately: built means I wrote the system; contributed as a developer means part of a larger team; provided advisory or data-quality support means diagnosis and recommendations; review means assessing work others did. Institutional ownership, operational acceptance and impact rest with the institution, and the linked records identify public projects rather than establishing sole authorship.

A question from Kampala

One project, several contracts.

At a Kampala workshop on 8 and 9 April 2026, an officer asked what happens when one project uses three procurement methods. That is a paraphrase, but the complication is the point: one project can connect to several contracting processes. Treating it as one contract loses those links.

If the portal forces those processes into one contract, people later lose the links they need to follow the project. The awkward case tests the design better than a tidy example.

The published account records 143 officers from 59 public bodies and 395 projects on the portal. The relationship between projects and contracting processes is explained in the OC4IDS publishing guidance.

A few working principles

The questions I keep asking.

Follow the connections.A record rarely answers the whole question alone. Look for the relationships that make it useful.
Test the awkward case.Ordinary work, missing information, and exceptions should shape the system, not be explained away.
Think about who inherits it.Ownership, maintenance, and handover belong in the design conversation from the start.

Professional context

Bring the question,
not a perfect brief.

I am based in Kampala and work as an Open Data Specialist with CoST. Infrastructure transparency is an important part of my experience, alongside the broader work of connecting data, software, and institutional needs.

If you are trying to publish information, understand its limits, or make different systems work together, I would be glad to hear what you are working through.

This is my personal account, not a statement on behalf of CoST. Any engagement begins by agreeing scope, authority, and capacity. You can also find my public code on GitHub.

Start a conversation See services