I'm Ben Kamau, a software engineer who builds backend systems that can't afford to be wrong.
I've architected five production platforms for banks, government programmes and financial-inclusion projects in Kenya, from multi-tenant core banking to a biometric registry for 2 million cattle. I lead teams from design to delivery, build the frontend when the product needs it, and I'm now building AI agents on the same foundations.
Case studies
The code is private and client names are withheld, so each study walks through the problem, the architecture, what I built and the results.
BodaPlus
A 14-service, multi-tenant core-banking platform for savings cooperatives serving motorcycle-taxi riders: onboarding, KYC, wallets, loans and welfare on an append-only ledger.
Mifugo360
Biometric identity for livestock: muzzle prints detected, embedded and matched by cosine similarity, giving close to 2 million cattle a unique digital ID.
E-Voucher
Digital vaccine-subsidy vouchers: farmers co-pay by M-Pesa, and an allocation engine splits every vaccination across nine stakeholder types and pays them through a partner bank.
Supply Chain Finance
Reverse factoring for a commercial bank: suppliers get paid early on approved invoices, with digital KYC, a configurable fee engine, USSD access and four-eyes approval throughout.
Dealer Finance
Working-capital loans for vehicle dealers at a licensed microfinance bank: three-step approval, tax-compliant fees, fine-grained authorization and a full audit trail.
Where I'm strongest
Backend & distributed systems
Java and Spring Boot microservices, gateways, service discovery, Kafka, transactional outboxes and multi-tenancy.
Systems that move money
Double-entry ledgers, loan engines, reconciliation, M-Pesa, and bank transfers: RTGS, EFT, B2B, B2C.
Security & identity
Keycloak, JWT and fine-grained access control; KYC against national identity, tax and credit-bureau systems.
Frontend & product
Backend-first, but I ship interfaces too: React and Next.js, designed around the APIs underneath. This site is one.
Data & ML
Python, pandas and data pipelines, plus running ML services in production behind a Java backend.
Technical leadership
Architecture, sprint planning and delivery with engineers, stakeholders and integration partners.
AI agents that can be trusted with money
Agents are moving out of demos and into real businesses: paying, approving, underwriting and reconciling on someone's behalf. The hard part isn't the model. It's everything around it: how an agent pays, what it's allowed to do, who it is, and the record of what it did.
That's the infrastructure I've spent years building for banks and government programmes. Now I'm building agents on top of it.
Where I'm focused
Where I want to be: at a startup putting agents into production where mistakes cost real money, whether that's fintech, agent infrastructure, or AI built for African markets.
- 01Agents inside regulated workflowsCredit, payments, reconciliation and compliance: work that's repetitive and rules-heavy, where an agent with the right guardrails saves real time.
- 02Tools agents can use safely (MCP)Exposing ledgers, KYC checks and bank APIs as tools with scoped permissions, spending limits, idempotency and approval gates.
- 03Agent paymentsStripe, Visa, Mastercard and others are building rails for agents that pay. I'm focused on how they meet M-Pesa and African banking.
- 04Testing before trustingEvaluating agent behaviour before it touches money, and logging every action so it can be explained afterwards.
- 05Shipping with agents every dayClaude Code and Codex are part of how I build. As community manager at Nairobi AI Community, I've run meetups and webinars on data science and AI since 2023.
Where I've worked
B.Sc. Actuarial Science, University of Nairobi, Second Class Upper. Data Science, Moringa School. A quantitative foundation for financial systems.