Edge AI Engineer
Edge AI engineers get machine learning models running directly on local devices — phones, IoT sensors, cameras, agricultural equipment — instead of relying on a constant connection to the cloud. This means aggressively optimising model size and speed (quantisation, pruning) so a device with limited compute and unreliable connectivity can still run useful AI in real time.
This is a particularly good fit for Kenya's realities: patchy rural connectivity means AI features for agriculture (crop-disease detection from a phone camera), security (offline surveillance analytics), and healthcare (diagnostic tools in low-connectivity clinics) need to work without depending on a live cloud connection.
- AI exposure
- 50 of 100, moderate exposure
- Hiring trend
- Growing
- Hiring rate
- 44%
- Minimum education
- Bachelor
The role
What the work is, what it pays, and what it costs you.
At a glance
- Remote friendly
- No
- Freelance potential
- Low
- Freelance rate
- Ksh 4,000
- Time to senior
- 5 years
A day in the role
"You learn to love a 2MB model that runs reliably offline more than a 200MB one that's marginally more accurate but needs a data connection nobody in the field actually has."
What it pays
Kenyan market, per month- Entry
- KES 120,000–190,000
- Mid
- KES 220,000–360,000
- Senior
- KES 390,000–630,000
The trade offs
In its favour
- Directly addresses Kenya's real connectivity constraints — high local relevance and impact.
- Deep, hardware-adjacent skill set that's hard to automate away.
Against it
- Smaller job market with fewer companies actively hiring for this specific niche.
- Often requires access to physical hardware for testing, which can slow iteration.
In practice
Deploy a working, fully offline computer-vision or audio model on a smartphone or Raspberry Pi — a concrete, working on-device demo carries far more weight than cloud-only ML projects for this role.
Progression runs ML engineer or embedded engineer → edge AI engineer → embedded AI platform lead, taking ownership of a company's on-device AI product line.
Agritech and off-grid IoT/solar hardware companies are the strongest local fit, given how directly this role addresses the country's connectivity and infrastructure realities.
A typical day includes optimising and benchmarking model variants for size/speed, testing on real target hardware, and debugging device-specific performance issues.
Exposure
How much of this a machine can already do, and how that was worked out.
Where this rating sits
1,516 rated careersRated above 67% of the 1,516 careers in the catalogue, which averages 43. Inside technology the mean is 62, across 125 careers.
What the rating is made of
Share of recorded tasks- Machine does it
- 20%Software can already complete this work end to end.
- Machine assists
- 45%A person still decides, but the drafting is done for them.
- Person does it
- 35%Judgement, relationships and accountability that do not transfer.
Named task by task
Already automated
- Generating quantisation configuration variants
- Benchmarking model size/speed trade-offs
Still human
- Compressing and optimising models to run within device compute/memory constraints
- Choosing hardware (microcontrollers, edge GPUs) appropriate for the use case
- Testing model accuracy degradation after compression
- Debugging performance issues specific to on-device deployment
Task counts
- Tasks recorded
- 8
- Automatable now
- 2
- Still human
- 4
- Augmenting
- Compression configuration testing,Benchmark automation
- Creating
- On-device AI product categories (offline diagnostics, offline crop monitoring)
Sources
Behind the rating- McKinsey State of AI 2025
Getting in
The routes into the role and what each one asks for.
What to study
8 courses- Certificate in Fashion Design and Textile TechnologyKsh 37,320a year
- Certificate in Desktop PublisherKsh 50,000a year
- Certificate in Mobile Applications and TechnologyKsh 56,420a year
- Certificate in Data Science and Artificial IntelligenceKsh 57,050a year
- Diploma in Photogrammetry and Remote SensingKsh 66,270a year
- Artisan in ICTKsh 67,189a year
- Certificate in Artificial Intelligence & CybersecurityKsh 67,189a year
- Certificate in Big DataKsh 67,189a year
How people get in
Computer/Electrical Engineering degree + embedded ML specialisation
5 years + 6 monthsMedium cost
Standard route combining embedded systems fundamentals with model-compression techniques.
ML engineer transition into embedded/edge deployment
6-12 monthsLow cost
Existing ML engineers add hardware-constraint-aware optimisation skills.
Certifications
Edge Impulse Certified Developer
Edge ImpulseKsh 01 months
Tools of the trade
TensorFlow Lite
AI/MLRequiredFree
Edge Impulse
AI/MLNice to havePaid
ONNX Runtime
AI/MLNice to haveFree
Who hires
Interview preparation
3 questionsHow would you shrink a model to run on a low-end Android phone without significantly hurting accuracy?
TechnicalMid
Look for discussion of quantisation, pruning, knowledge distillation, and careful accuracy-vs-size benchmarking rather than guesswork.
Design an offline crop-disease detection feature for a phone with no data connection. What are the key constraints?
SituationalMid
Should address model size, on-device inference speed, battery impact, and how the app handles eventual sync when connectivity returns.
What's the difference between quantisation and pruning?
TechnicalEntry
Quantisation reduces numerical precision of model weights; pruning removes less-important weights/connections entirely. Good candidates know both are often combined.
Common misconceptions
It's the same job as regular ML engineering, just on a smaller device.
Edge deployment requires genuinely different techniques — quantisation, pruning, hardware-specific optimisation — and a different mindset around strict resource constraints.
You always need custom hardware to do this well.
A lot of edge AI work runs on standard smartphones using efficient model formats (TensorFlow Lite, ONNX), no exotic hardware required.
What happens next
How the role changes from here, and where it leads.
The near term
Well-suited to Africa's connectivity realities, with growing hardware-adjacent demand
- More affordable edge-capable chips lowering the cost of deployment
- Growing product categories in offline agri/health diagnostics
- What to do
- Build and deploy a real on-device model (even on a smartphone) that works fully offline — demonstrating the specific optimisation skill this role needs.
Where pay is heading
2024 to 2030Monthly pay in Kenyan shillings, rounded to the nearest thousand. These are projections, not observations.
Growth outlook
- Net demand change
- 20
- Over
- 2025-2028
- Drivers
- Growth of AI-enabled IoT and mobile products for low-connectivity markets,Falling cost of edge-capable hardware
- Headwinds
- Smaller, more specialised talent pool and slower hiring pipelines
Supply and demand
- Demand
- 48
- Supply pressure
- 30
- Balance
- Balanced
What to learn
- Model quantisation and pruning
- Embedded systems programming
- Hardware-aware model architecture design
Tools worth knowing
TensorFlow Lite
Priority: Essential
Mobile/embedded model deployment
Edge Impulse
Priority: Recommended
End-to-end edge ML development platform
Where people move next
3 recorded movesLine length under each name is the distance of the move: shorter means more of what you already do carries over. Marked lines are steps up rather than sideways.
- Ai Ml Engineer
Easy60% skill overlapLateral
Shared ML foundations, generalising beyond edge-specific constraints.
- Iot Developer
Easy55% skill overlapLateral
Overlapping hardware/embedded-systems skill set.
- Robotics Engineer
Moderate40% skill overlapPromotion
Extends embedded AI skill into full robotic systems.
Related careers
Kenyan market notes
Agritech (crop/livestock monitoring from mobile cameras) and off-grid solar/IoT hardware companies are the strongest local fit, given how directly edge AI addresses Kenya's connectivity constraints.
Further reading
This role is rated 50 out of 100 today. Save it and the app keeps that number, then tells you by how much it has moved when the record is next reviewed.