We build AI that stays running.
B&G Enterprises is a Nairobi-based AI systems company. We’ve been doing machine learning and data infrastructure work since before “AI” became a sticker on every product page. We focus on systems — the unglamorous parts that keep working at 3am.
Demos are easy. Production is the real work.
We believe the AI field has a demo problem. A beautiful prototype wins applause; a system that runs for a year in a bank, insurer, or logistics firm wins trust. We pick the second one every time.
We work with operators, not pilots. Our engagements are priced in outcomes, scoped in weeks, and graded on what’s still running six months later. The boring parts — eval harnesses, monitoring, rollback, cost guards, human-in-the-loop — are not afterthoughts. They are the work.
We refuse projects we can’t instrument. We refuse deliverables we can’t explain. We refuse scope that treats AI as a magic layer sprinkled over existing problems, rather than a capability that respects the process, the people, and the data.
Six principles that shape every engagement.
Boring is a feature.
Robust, measurable, and auditable systems outrank clever ones. Every shortcut taken in year one is paid back with interest in year two.
Start where the evidence is.
We look at your data, your processes, and your ops before we look at models. Most AI failures are data or workflow failures wearing an AI costume.
Humans stay in the loop.
High-stakes decisions get explicit reviewer workflows, approval gates, and audit trails. Automation augments judgement; it doesn’t replace accountability.
Measure or don’t ship.
Every model has eval suites. Every system has dashboards. Every rollout has a rollback. If we can’t measure it, we don’t promise it.
Own the full stack.
Data, models, orchestration, APIs, UI, observability — we take responsibility for the whole loop, not just the ML bit. The handoff is where projects die.
Context is East Africa.
We design for M-Pesa, WhatsApp, bandwidth realities, regulatory regimes (CBK, IRA, ODPC), and the actual teams running these systems in Nairobi, Mombasa, Kampala, and Kigali.
Who you’ll actually work with.
We don’t hand you off to juniors after the sales call. Founders are on every engagement.
Background in applied ML, data platforms, and production systems. Led AI engineering at regional fintech and logistics firms before co-founding B&G. Focus areas: model deployment, evaluation harnesses, data pipelines, vector & RAG systems.
Believes the only AI worth building is AI you can show the numbers on. Has opinions about feature stores.
Product and operations background across insurance, health, and public-sector work. Runs discovery, scoping, stakeholder alignment, and delivery. Obsessed with the gap between “demo” and “in production” — specifically, the ops, governance, and change-management work that closes it.
Enjoys writing SOWs that survive contact with a legal team. Long experience with regulated industries.
How we got here.
Founded in Nairobi.
Started as a two-person data & ML consultancy focused on financial-services work — credit scoring, M-Pesa transaction analytics, and the messy reconciliation problems that were, and still are, everywhere.
First production deployments.
Fraud scoring for a regional payment processor. Claim-triage model for an insurer. Forecasting for a logistics firm. Learned, fast, that the models were the easy part — the integration with legacy systems, reviewer workflows, and monitoring was 80% of the work.
Productised the methodology.
Turned the recurring patterns into three engagement models (Audit, Integration, Transformation) and eight capabilities. Standardised the eval harnesses, the observability stack, and the human-in-the-loop patterns we kept re-implementing.
LLM and agent era.
Began shipping RAG systems, conversational agents on WhatsApp and IVR, and workflow-automation agents. Learned, again, that the hard parts are retrieval quality, tool discipline, and cost controls — not the model itself.
Regulatory & governance depth.
Extensive work with CBK, IRA, and ODPC alignment. Published internal playbooks for model risk, fairness, privacy, and audit-readiness. Became the firm that the compliance teams want in the room, not the one they want blocked.
Operators, not pilots.
We work with a small number of clients at a time, deeply. Every engagement is scoped against outcomes, measured with evidence, and graded on what’s running months after we’ve stepped back. That’s the whole business.
Built in Nairobi. Serving East Africa.
We’re based in Nairobi, Kenya. Most of our work happens across East Africa — Kenya, Uganda, Tanzania, Rwanda — and occasionally further afield when a partner engagement pulls us.
Being in Nairobi matters. It means we design for the actual context: M-Pesa-first payments, WhatsApp-native customers, bandwidth-constrained deployments, CBK and IRA regulatory regimes, ODPC data-protection reality, and the specific labour markets for data, ops, and support teams.
It also means we’re accessible. Same time zone. Same working week. In-person workshops when they add value. You don’t have to schedule meetings at 11pm local time to talk to us.
production AI
8 sectors
critical systems
running at 12 months
Interested in working together?
Sixty minutes on a call. Bring a problem, a dataset, or just a question. If we’re not the right fit, we’ll say so and point you somewhere better.