Portfolio

Systems we’ve built.

Real products across geospatial, software and hardware. Each one started as a concrete problem and grew into a working system.

Navigation · Geospatial

SalamaNav — safe navigation for Kenyan roads

A driver-assistance platform built for Kenyan conditions: real-time hazard alerts, offline-first navigation and a prepaid M-PESA wallet.

The challenge

Drivers contend with unmarked potholes, speed cameras and checkpoints, patchy connectivity outside the cities, and navigation apps that assume an always-on connection and a credit card.

The build

A .NET 10 backend over PostgreSQL/PostGIS, an Android client with offline map caching, and a React web dashboard. Hazards are crowdsourced and verified, routes are priced pay-per-use, and the wallet tops up through M-PESA.

What it delivers

Turn-by-turn navigation that keeps working offline, community-sourced hazard warnings, and a prepaid model that fits how Kenyan drivers actually pay.

Navigation PostGIS M-PESA Offline-first

Fleet · Geospatial

Wayrivo — fleet intelligence in motion

A connected-mobility platform for GPS tracking and fleet operations: telemetry, visibility, billing, notifications and partner APIs in one system. In active testing.

The challenge

Fleet operators run GPS hardware but have no unified way to turn raw device telemetry into live visibility, compliance, billing and partner integrations.

The build

A .NET service mesh split into GPS data management (receiver and processor for GT06 and Navtelecom frames), fleet API, organization management, payment-billing, notifications and an external API platform, fronted by platform-console, customer, partner and developer portals.

What it delivers

Live fleet visibility, normalized device telemetry, org management, billing and a governed external API, kept honest by CI gates on the GPS listener and API contracts.

In testing GPS Tracking Telematics .NET Fleet Ops

IoT · AI · Hardware

EdgeSentience — AI-powered IoT camera platform

An end-to-end surveillance camera system: ESP32-CAM devices, cloud media processing and a real-time monitoring dashboard.

The challenge

Turning low-cost ESP32-CAM modules into a dependable surveillance product, with secure device onboarding, reliable frame upload, motion detection and no manual provisioning.

The build

Three .NET 10 services (device control plane, media ingest, dashboard API) over PostgreSQL and Redis, a React 19 dashboard, and ESP-IDF firmware with OTA updates and a captive-portal WiFi setup. Footage lands in OmnixStorage under retention policies.

What it delivers

Live and historical camera feeds, motion detection, remote OTA firmware updates and JWT-authenticated device access, on commodity hardware.

ESP32-CAM .NET 10 React Motion Detection

Infrastructure · Software

OmnixStorage — S3-compatible object storage

A high-performance, distributed object storage system built in .NET 10, with a drop-in S3 API.

The challenge

Building object storage that existing S3 tooling can point at without change, while keeping throughput, durability and cluster operation simple.

The build

io_uring for high-throughput I/O, SIMD Reed-Solomon erasure coding for durability, Raft consensus for metadata and SWIM gossip for cluster health, behind AWS SigV4 authentication and an IAM policy engine.

What it delivers

An S3-compatible endpoint your tooling already speaks, with native erasure coding and self-organizing cluster membership.

S3 Erasure Coding Raft .NET 10

Solar · Software

SunAlchemy — solar system sizing, done rigorously

A technical solar-plus-battery sizing tool that turns roof data, appliance loads and inverter surge requirements into one installer-ready report.

The challenge

Solar designers were sizing systems with rough rules of thumb, producing proposals that didn’t survive engineering review.

The build

A web tool that combines roof data, appliance load profiles, battery sizing, inverter surge requirements and equipment assumptions into a single defensible calculation.

What it delivers

An installer-ready technical report, so proposals stop guessing and start from aligned, reviewable numbers.

Solar Battery Storage Sizing Web App

Software · AI

Niongelelee — text-to-speech that turns words into audio

A text-to-speech progressive web app with upfront pricing, job tracking and a prepaid wallet.

The challenge

Making high-quality speech synthesis usable as a paid service, with transparent pricing, reliable job tracking and a payment method that matches local mobile-money habits.

The build

A .NET 9 Web API and React/TypeScript frontend. Text is priced up front (50 KES per 200 words), queued as a background job, synthesized with retry logic, and served back for playback or download. Hangfire drives the queue; the wallet tops up through M-PESA.

What it delivers

A progressive web app that turns typed text into downloadable MP3 audio, with real-time job status via SignalR and pay-per-use pricing.

TTS React M-PESA PWA

On the roadmap

Capabilities we’re building toward.

Target profiles for systems we intend to build next. Each one is a direction we’re working to turn into a real, delivered project.

Logistics · Geospatial

Fleet telematics platform

Live GPS tracking, route optimization and operational dashboards for a moving fleet.

The challenge

Operators had telemetry but no single view: where each vehicle is, whether routes are optimal, and where fuel and idle time leak.

The build

An ingestion service for GPS feeds, PostGIS for live positions, and BI dashboards over the same data.

The outcome

[Client] cut deadhead miles by [X%] in the first [N] months.

Fleet GIS BI

Infrastructure · 3D

Digital twin for infrastructure

A living 3D model fused with IoT sensor streams and GIS context.

The challenge

Static renders went stale the day they were delivered. Operators needed the asset to reflect reality, not a snapshot.

The build

A 3D model anchored to real coordinates, IoT sensors streaming telemetry, and GIS layers for surrounding context.

The outcome

[Asset owner] now watches live load and temperature against the model, and flags anomalies before they become failures.

Digital Twin IoT 3D

Defense · GEOINT

Satellite imagery analysis pipeline

Automated feature extraction over imagery, with a human in the loop for the final call.

The challenge

Analysts were drowning in imagery; the valuable detections were a fraction of the volume.

The build

An extraction pipeline that pre-flags changes, plus a review interface where an analyst confirms or rejects each one.

The outcome

[Agency] analysts triage [N] km² of imagery a day instead of sifting frame by frame.

GEOINT AI Imagery

Product · Hardware

IoT sensor device, board to enclosure

A connected sensing device taken from circuit sketch to manufactured part.

The challenge

A working idea needed to become a durable, manufacturable product: electronics, firmware and enclosure together.

The build

A custom PCB laid out in KiCad, firmware in C, a 3D-designed enclosure, and CNC or additive production for the first units.

The outcome

[Company] moved from prototype to a first run of [N] field units in [N] weeks.

PCB Firmware CNC / 3D Print

These are illustrative capability profiles, not claims about delivered work. Every [bracketed] detail is a placeholder: swap it for a real project, with the client’s permission, before publishing.

Want your project to be next?

Every system here started as a conversation about a problem. Tell us yours.

Request a Quote →