RAG / Retrieval Engineer
RAG (retrieval-augmented generation) engineers build the systems that let an LLM answer questions grounded in a company's own documents and data, instead of relying only on what the model memorised during training — retrieving relevant chunks of text or data at query time and feeding them into the model's context. This is how most 'chat with your documents' and internal knowledge-assistant products actually work under the hood.
Almost every Kenyan company building an internal AI assistant — for HR policy questions, legal document search, or customer support grounded in product documentation — needs this exact skill, making it one of the more immediately practical and widely-applicable AI specialisations to learn.
- AI exposure
- 57 of 100, moderate exposure
- Hiring trend
- Growing
- Hiring rate
- 70%
- Minimum education
- Bachelor
The role
What the work is, what it pays, and what it costs you.
At a glance
- Remote friendly
- Yes
- Freelance potential
- High
- Freelance rate
- Ksh 4,800
- Time to senior
- 4 years
A day in the role
"Most bugs I fix aren't in the model — they're in retrieval. If you feed the LLM the wrong document chunk, no amount of prompting fixes the answer."
What it pays
Kenyan market, per month- Entry
- KES 130,000–200,000
- Mid
- KES 240,000–380,000
- Senior
- KES 400,000–650,000
The trade offs
In its favour
- Extremely broad applicability — nearly every company with documents can use this skill.
- Fast-growing, well-compensated, with a relatively approachable learning curve for backend engineers.
Against it
- Getting genuinely good retrieval quality on messy real-world data is harder than tutorials suggest.
- Growing model context windows may reduce demand for RAG in some simpler use cases over time.
In practice
Build a RAG system over a genuinely messy real document set (e.g. scanned PDFs, inconsistent formatting) rather than a clean demo dataset — this is what separates a hireable portfolio project from a tutorial follow-along.
Progression runs backend engineer → RAG engineer → AI platform engineer, taking on ownership of the retrieval and knowledge-grounding layer across multiple internal AI products.
Nearly every mid-to-large Kenyan company building an internal AI assistant needs this skill, making it one of the most immediately employable AI specialisations available today.
A typical day includes tuning chunking/retrieval configuration, debugging specific query failures, and reviewing evaluation metrics for answer accuracy and groundedness.
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 78% 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
- 25%Software can already complete this work end to end.
- Machine assists
- 50%A person still decides, but the drafting is done for them.
- Person does it
- 25%Judgement, relationships and accountability that do not transfer.
Named task by task
Already automated
- Generating document-chunking configuration variants
- Drafting retrieval-quality test cases
Still human
- Designing document chunking and embedding strategies for a specific knowledge base
- Tuning retrieval relevance and ranking quality
- Debugging cases where the model 'hallucinates' despite having correct source documents retrieved
- Building evaluation pipelines for answer accuracy and groundedness
- Deciding when to use RAG versus fine-tuning versus a longer context window
Task counts
- Tasks recorded
- 10
- Automatable now
- 2
- Still human
- 5
- Augmenting
- Chunking configuration testing,Evaluation test-case generation
- Creating
- Retrieval evaluation tooling,Hybrid search architectures
Sources
Behind the rating- Stanford HAI AI Index 2025
- 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
Software Engineering degree + RAG framework self-study
4 years + 3-6 monthsMedium cost
Standard degree, then hands-on practice with LangChain, LlamaIndex, and a vector database.
Backend engineer transition
3-6 monthsLow cost
Fastest route for experienced backend engineers — the core skill (data pipelines + API integration) transfers directly.
Certifications
DeepLearning.AI Building and Evaluating Advanced RAG
DeepLearning.AIKsh 01 months
Tools of the trade
LlamaIndex
AI/LLMRequiredFree
LangChain
AI/LLMRequiredFree
Pinecone
DatabaseNice to havePaid
Weaviate
DatabaseNice to haveFree
Who hires
Interview preparation
3 questionsA RAG system retrieves the right document but the LLM still gives a wrong answer. How do you debug this?
TechnicalMid
Look for checking chunk size/context window fit, whether the retrieved chunk actually contains the answer clearly, and prompt instructions telling the model to cite/ground its answer.
How would you choose between RAG, fine-tuning, and simply using a longer context window for a given use case?
TechnicalSenior
Should weigh factors: how often the underlying data changes (RAG wins for frequently-updated data), cost, latency, and need for source citation.
How do you evaluate whether your retrieval system is actually working well?
TechnicalMid
Look for a real evaluation set with known correct answers, metrics like retrieval precision/recall, and end-to-end groundedness scoring, not just eyeballing a few examples.
Common misconceptions
RAG completely eliminates hallucination.
It significantly reduces it by grounding answers in real documents, but a model can still misread or misrepresent retrieved content — evaluation and groundedness-checking remain essential.
It's just plugging a vector database into a chatbot.
Getting retrieval quality genuinely good requires careful chunking strategy, embedding model choice, re-ranking, and continuous evaluation against real user queries.
What happens next
How the role changes from here, and where it leads.
The near term
Near-universal building block for enterprise AI assistants
- Hybrid search (keyword + semantic) becoming standard practice
- Growing focus on citation/groundedness evaluation as a trust requirement
- What to do
- Build a real RAG system over a genuinely messy document set (not a clean tutorial dataset) — the hard, hireable skill is handling messy real-world retrieval, not the happy path.
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
- 40
- Over
- 2025-2028
- Drivers
- Every company with internal documents wants a grounded AI assistant,Enterprise trust requires citation-backed, verifiable AI answers
- Headwinds
- Growing context windows in newer models reducing RAG's necessity for some use cases
Supply and demand
- Demand
- 88
- Supply pressure
- 40
- Balance
- High demand
What to learn
- Vector search and embeddings
- Chunking and retrieval evaluation
- Hybrid search (keyword + semantic)
Tools worth knowing
LlamaIndex
Priority: Essential
RAG pipeline construction
Pinecone
Priority: Recommended
Managed vector database for retrieval
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.
- Agentic Ai Developer
Easy65% skill overlapLateral
Natural extension into multi-step agent systems that use retrieval as one tool among several.
- Vector Database Engineer
Easy60% skill overlapLateral
Specialises deeper into the retrieval infrastructure layer specifically.
- Ai Ml Engineer
Moderate45% skill overlapPromotion
Broadens from application-layer retrieval into model training/fine-tuning.
Related careers
Kenyan market notes
One of the most immediately in-demand AI specialisations locally, since almost any company with internal documents wants an AI assistant grounded in them — legal, HR, customer support, and product documentation are the most common use cases.
Further reading
This role is rated 57 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.