The Astexlabs Flywheel: Services, Research, Products

title: "The Astexlabs Flywheel: Services, Research, Products" date: "2026-01-05" description: "How we're building a sustainable research lab by spinning client problems into proprietary datasets and Micro-SaaS products." author: "Astexlabs Team" tags: ["Business", "Strategy", "Research"] published: true

Most African tech companies fall into one of two camps:

Service shops that grind on client work with no IP
Startups that burn cash on products nobody needs

Astexlabs is different. We operate a flywheel model that converts client revenue into research assets and scalable products.

The Three Engines

1. Service Engine (Revenue)

We take on high-complexity engineering contracts:

Fintech platforms that must handle unreliable banking APIs
Logistics systems that work with messy, unstructured addresses
Energy management tools that operate offline during power outages

We don't build brochure websites. We build systems that must work when everything else fails.

Why this matters: These projects generate immediate cash flow while exposing us to real-world African tech problems.

2. Research Engine (Asset Generation)

Every client project yields data:

From Fintech Work: Corpus of Nigerian bank alert formats
From Logistics Work: Dataset of Lagos descriptive addresses
From User Interviews: Code-mixed Pidgin-English transcripts

We anonymize, structure, and store this data. Over time, we build proprietary datasets that:

Improve AI model performance on African contexts
Provide competitive moats for our products
Serve as sellable assets to global AI companies

Why this matters: We're not just billing hours—we're building IP.

3. Product Engine (Scale)

We identify patterns in client problems and spin them into Micro-SaaS products:

Example 1: BankAlertParser API

Problem: Every fintech client needs to parse Nigerian bank SMS alerts
Solution: API that converts messy SMS into structured JSON
Market: Nigerian fintech startups, accounting software, expense trackers

Example 2: NaijaGeocoder

Problem: Google Maps fails on descriptive Lagos addresses
Solution: API that converts "beside the yellow mosque" into lat/long
Market: Logistics, delivery, ride-hailing apps

Example 3: PidginNLP SDK

Problem: Chatbots don't understand code-mixed Nigerian English
Solution: NLP library fine-tuned on Pidgin and code-switching
Market: Customer service platforms, voice assistants

Why this matters: Products scale beyond our time. We sell the same solution 1,000 times instead of building custom once.

The Flywheel Effect

Client Project → Problem Discovery → Dataset Collection
                                    ↓
Product Revenue ← Product Launch ← Research Insights
       ↓
More R&D Budget → Better Products → More Clients

Each cycle strengthens the next:

More clients = more data = better research
Better research = better products = more revenue
More revenue = hire better engineers = take on harder problems

Why This Works in Africa

The African tech ecosystem suffers from a "Missing Middle":

Bottom: Generic outsourcing shops (no differentiation)
Top: Well-funded startups (high burn rate, unclear paths to profitability)
Middle: Research-driven engineering labs (basically non-existent)

We occupy that middle ground. We're:

Profitable (from services)
Research-driven (building proprietary datasets)
Product-focused (spinning IP into scalable revenue)

The Inspiration

We model ourselves after:

Anysphere (Cursor): Building tools to revolutionize workflows
Anthropic: Prioritizing safety and cultural alignment in AI

But adapted for the African context: lower capital intensity, faster revenue cycles, real-world problem validation.

Current Traction

Services: 5 active client projects (fintech, logistics, energy)
Research: 3 datasets in development (bank alerts, addresses, Pidgin NLP)
Products: 2 in private beta (BankAlertParser API, NaijaGeocoder)

We're not VC-funded. We're sustainably building the kind of company Africa needs—one that generates both revenue and intellectual property.

Want to work with us? hello@astexlabs.com