Augusta R&D

We explore
so you don't have to.

Augusta invests 10% of revenue into a working R&D programme — applied AI, edge computing, IoT firmware, low-bandwidth UX. We do the early experiments so when a customer asks something hard, the answer is "we already built a prototype."

Read our R&D field notes
10%
Revenue invested in applied R&D each year
14
Active research tracks across applied AI, IoT, and infrastructure
8
Engineers on the dedicated R&D team
23
R&D outputs upstreamed into customer engagements in 2025
The mandate

Research has to reach production.

Augusta's R&D isn't an academic activity. Every research track has a hypothesis, a budget, a sponsor on the leadership team, and a six-month review where it either continues, pivots, or shuts down.

The work is judged by one question: did it reach a paying customer?

Our R&D programme exists because the work we do for customers gets harder every year. Five years ago, "AI" was a feature. Today, it's table-stakes. Five years from now, we expect it'll be agentic systems making operational decisions inside the platforms we ship.

"R&D isn't where we play. It's where we discover what production looks like in 18 months — and start being able to deliver it now."

Augusta R&D operates as a small, dedicated team of senior engineers and architects, separated from delivery so they can take the long view. They publish internally, run weekly demos for the engineering team, and spike out prototypes that delivery teams can pick up when a customer brief calls for them.

The team is led by Titus Wangusi, who reports directly to Alexander Pedro, our CTO.

Active tracks

What we're currently working on.

Six of our fourteen active research tracks. The full list is shared with customers under NDA on request.

01 / Applied AI

Agentic workflows for back-office

Production-safe agent orchestration for finance, procurement, and operations workflows. RAG over enterprise documents, structured tool use.

In production · 3 customers
02 / Edge / IoT

Solar-powered telematics

Ultra-low-power firmware for off-grid asset tracking — months of operation on a small solar panel without supervision.

Prototype · field testing
03 / Mobile

Local-first architecture

CRDT-based mobile sync that survives multi-day disconnections. Construction sites, field surveys, last-mile dispatch.

Production · open-sourced
04 / Data

Multilingual NLP for Swahili

Production-quality Swahili and Sheng entity extraction, sentiment, and intent classification — for customer service and compliance.

Beta · 2 pilot customers
05 / Infrastructure

Cost-optimised cloud for Africa

FinOps tooling tuned for AWS af-south-1 / Azure South Africa workloads. Reserved-capacity strategy at the East African scale.

Internal tool · selective rollout
06 / Security

DPA-aware data engineering

Tooling for automatic PII detection, retention enforcement, and data-subject-request handling across data warehouses.

Prototype · internal use
How R&D works

Four-stage funnel from idea to production.

Most ideas get killed at stage two. The ones that survive are battle-tested before they hit a customer.

01

Hypothesis

Written one-pager — what we'd build, why it'd matter, what could disprove it.

02

Spike

2–4 week timeboxed prototype. Either it shows promise or we kill it cleanly.

03

Internal pilot

Used inside Augusta — by our engineers, on our own work — to harden it.

04

Customer release

Carefully sequenced rollout to customer engagements where it's the right fit.

2025 outputs

What R&D shipped last year.

Six concrete outputs from the 2025 R&D programme that found their way into production.

01

MwalimuPLUS adaptive engine

The personalisation algorithm now serving 200,000+ Kenyan students started life as a six-week R&D spike on adaptive curriculum sequencing.

02

Speed-limiter telemetry compression

A novel encoding for high-frequency vehicle telemetry that cut our data-egress bills by 38% without losing forensic detail.

03

Internal codegen agent

An LLM-driven internal tool that generates ADR drafts, runbooks, and post-mortem reports from raw incident data — used on every Augusta engagement.

04

M-Pesa reconciliation library

Open-sourced reconciliation library that's now the de-facto reference implementation for Daraja-based payment platforms in East Africa.

05

Offline CRDT toolkit

Local-first sync library shipped into the Hampton Construction mobile app — and made available to other Augusta projects as an internal package.

06

Augusta IMPACT methodology v3

The third iteration of our delivery methodology, refined from the operating data of 50+ engagements and now in use across all customer projects.

Curious?

Working on something research-heavy?

If you're exploring something at the edge of what's been done — applied AI, novel hardware, unusual data infrastructure — we'd be interested in talking. Some of our most rewarding engagements start as joint research.

R&D contact

Lead
Titus Wangusi, R&D Manager