In-house builds, and everything else.

Two tracks of work, one standard. The first five are in-house systems built for the Kenyan market. Everything after is client and range work: marketing sites, e-commerce, POS, analytics, security, design and whatever comes next. Click any client project to open its showcase.

Track 01 · In-house

Flagship systems · Kenya

Five production builds made for the local market. Real users, real revenue, real outcomes — the case studies that show how I think.

5 systems in production Track 02 · Client & range

Client & range work

Marketing sites, e-commerce, portfolios, POS, analytics, security and web design. Click any card to open its showcase modal.

0 projects and counting
01 · IN-HOUSE

Flagship builds

5 systems · Kenya
What's our refund window for enterprise plans?
Enterprise plans carry a 30-day refund window from the invoice date.
Policy 4.2MSA §7.1
AI · RAG01 / 05

Sauti — RAG knowledge assistant

A retrieval-augmented assistant built over private company documents — policies, contracts, PDFs and internal wikis. Every answer cites the exact source it came from, so trust is verifiable instead of assumed.

92%Answer accuracy
<2sMedian response
−78%Repeat questions
PythonFastAPILangChainpgvectorOpenAI
View case study AI · RAG · Python
KES 4,850
REF · PF-04812
Send
Payments02 / 05

PesaFlow — M-Pesa checkout & auto-reconciliation

A production Daraja integration: STK Push at checkout, idempotent callbacks that survive retries, and a reconciliation engine that matches every payment to its order.

99.4%Callback success
0 hrsManual recon / week
2.7sPIN to paid order
Daraja APINode.jsPostgreSQLRedis
View case study Payments · M-Pesa
Revenue
2.41M
Orders
1,284
Returns
3.2%
Data & BI03 / 05

Sokoni Insights — data pipeline & live dashboard

A scheduled ETL pipeline pulling from POS, M-Pesa, ops sheets and APIs into one clean model — with validation that blocks bad data before it reaches a decision-maker.

6 hrsReporting removed / wk
DailyData freshness
1Source of truth
PythonAirflowdbtPostgreSQLMetabase
View case study Data · ETL · BI
New WhatsApp messagetrigger
Classify intent + urgencyai
Create ticket · reply · logaction
AI · Automation04 / 05

Rafiki — AI workflow automation engine

Takes messy inbound — WhatsApp, email, forms, tickets — classifies intent, calls the tools it needs, and routes the outcome. High-confidence runs proceed automatically.

74%Handled without a human
~40hTriage removed / month
5.2sMessage to action
TypeScriptNode.jsOpenAIWhatsApp APIBullMQ
View case study AI · Workflows · Agents
app.kazidesk.co.ke/jobs
Achieng' Plumbing · leak repairActive
Tumaini Cleaning · deep clean
Mwangaza Electrical · rewiringScheduled
Product · SaaS05 / 05

KaziDesk — micro-SaaS for service teams

A lightweight job and client desk for small service businesses — plumbers, electricians, cleaners, fundis. Idea to first paying customer in eight weeks.

KES 84kMonthly recurring revenue
23Paying teams
4.2%Monthly churn
Next.jsTypeScriptPrismaM-PesaStripe
View case study Product · Micro-SaaS
02 · CLIENT & RANGE

Client & range work

loading…

Everything shipped for clients or as standalone range work — marketing sites, e-commerce, portfolios, POS, analytics, security and design. Click any card to open its showcase. Search by name, technology or category.

Showing 0 of 0 projects

Want something like this built for you?

I work with local and remote clients on freelance and contract engagements. Tell me the problem, the timeline and what success looks like — I'll reply with an honest take on whether I'm the right fit.

Start a conversation Back to home