From Silicon Savannah to an AI Powerhouse: Kenya’s Leading AI Firms

AgriSense (Agri‑AI)

AgriSense applies computer vision and satellite data to farm-level decision tools for smallholder crops. Their product combines weekly satellite indices, soil-moisture models and mobile SMS alerts to support planting and input scheduling. Typical customers are agribusinesses and NGOs running farmer-extension programmes. Suggested research sources: company website, LinkedIn team pages, TechCabal/Disrupt Africa articles, accelerator demo days, public pilot announcements from NGOs or telco partners.

• Focus: smallholder yield forecasting, pest detection, weather-smart alerts
• Deployments: field pilots with cooperatives or extension services (verify sources)
• Verification sources: company site, LinkedIn, local press, accelerator cohorts

MobiCredit AI (Fintech risk & scoring)

MobiCredit AI builds alternative credit-scoring models that ingest mobile-transaction features, utility payments and psychometric data. Their models are designed to work with sparse formal credit histories common across East Africa and to integrate with lender APIs for decisioning. Useful for banks and digital lenders exploring thin-file customers. Suggested research sources: product demos, LinkedIn, fintech accelerators and local regulatory filings.

• Focus: alternative credit scoring, lender risk pipelines
• Typical customers: digital lenders, microfinance institutions
• Verification sources: company website, fintech press, LinkedIn

ClinicAI (Healthtech triage & workflow)

ClinicAI focuses on clinical triage models and workflow orchestration for primary-care clinics. The platform standardises intake, routes likely cases to telemedicine or referral and aggregates anonymised clinical signals for district-health planning. Emphasis is on low-bandwidth user interfaces and compliance with local data protection rules. Suggested verification: clinic pilot press releases, NGO partners, Ministry of Health pilot lists and GitHub or technical documentation when available.

• Focus: triage, referrals, clinic workflow optimisation
• Design constraints: offline-first, low-bandwidth UIs
• Verification sources: pilot announcements, LinkedIn, health sector reports

LocalLang Labs (NLP & local languages)

LocalLang Labs builds language models and NLP pipelines tailored to Kiswahili, Sheng and other regional dialects. Core product modules include intent classification for chatbots, automated content moderation and speech-to-text for low-resource languages. They collaborate with contact centres and civic tech organisations to localise user experiences. Suggested sources: GitHub repos, academic collaborations, company blog, local tech coverage.

• Focus: Kiswahili and dialect NLP, speech-to-text and NLU
• Use cases: customer support chatbots, civic engagement tools
• Verification sources: GitHub, research partnerships, LinkedIn

FarmLedger (Supply-chain traceability)

FarmLedger combines lightweight sensor data, mobile input records and simple ML models to track produce from farm to buyer, improving traceability and quality signals for exporters and processors. Their product emphasises interoperability with existing aggregator apps and buyer dashboards. Verify via buyer pilot announcements, export chain partners and local agri-tech press.

• Focus: traceability, quality prediction, buyer dashboards
• Customers: aggregators, exporters, processors
• Verification sources: press, aggregator websites, LinkedIn

GovAssist AI (Public-sector automation)

GovAssist AI offers document-intake, citizen-service routing and simple automated case triage for local governments. Designed to integrate with government digital services and citizen call centres, the product supports rapid prototyping for 311-style services and benefits distribution. Suggested verification: government pilot listings, procurement notices and accelerator demo day coverage.

• Focus: document automation, citizen-service routing, case triage
• Design: integration with existing government portals and call centres
• Verification sources: procurement portals, local press, LinkedIn

MedData Secure (Privacy-aware health AI)

MedData Secure provides privacy-aware data pipelines for health research and ML model training, using de-identification patterns and role-based access suited to Kenyan data protection expectations. The company focuses on enabling safe model development for NGOs and research labs without exposing identifiable records. Suggested verification: technical whitepapers, research partnerships, NGO announcements and privacy compliance materials.

• Focus: secure data pipelines, de-identification, governed model training
• Clients: research networks, NGOs, private clinics
• Verification sources: whitepapers, research partners, LinkedIn

RetailIntel KE (Retail and agent analytics)

RetailIntel KE builds demand-forecasting and assortment-optimisation tools for FMCG supply chains and agent networks. The product works with POS data, agent sales logs and low-bandwidth reporting to suggest inventory rebalancing and promotional tactics. Suggested verification: retailer case studies, distributor partnerships and local trade press.

• Focus: demand forecasting, assortment optimisation for FMCG
• Data inputs: POS, agent sales and distributor logs
• Verification sources: retailer case studies, LinkedIn, trade press

ClimateEye (Environmental monitoring)

ClimateEye uses remote sensing and local sensor fusion to provide flood risk alerts, land-use change detection and water-stress indices for agricultural and municipal users. Outputs are provided as mobile alerts for community responders and dashboards for planners. Verify via NGO pilots, satellite-data sources and university partnerships.

• Focus: flood alerts, water-stress monitoring, land-use change detection
• Users: municipalities, NGOs, agribusinesses
• Verification sources: academic partners, local press, LinkedIn

VoiceEnroll (Identity & voice biometrics)

VoiceEnroll offers voice-based identity verification and IVR automation for low-literacy user bases, optimised for phone networks with intermittent connectivity. Target customers include telco agents, micro-lenders and government services requiring remote identity confirmation. Suggested verification: pilot notices, telco partnerships, GitHub demos and regulatory filings where applicable.

• Focus: voice biometrics, IVR automation for low-literacy contexts
• Typical deployments: telcos, lenders, government service hotlines
• Verification sources: pilot press, LinkedIn, demo repositories

Exportable CSV columns

Use these CSV columns to import the timeline into spreadsheets or BI tools.

• year,event,type,source
• 2010,"iHub launch","hub_launch","iHub historical page"
• 2019,"Kenya Data Protection Act","regulatory","Government publication"

Contract clauses checklist

Key clauses to include in PoC and procurement contracts to reduce legal and operational risk.

• Data access & use: permitted purposes, retention periods, anonymisation requirements, allowed downstream processing
• IP boundaries: pre-existing IP, PoC-created IP ownership and licensing, rights to models and weights
• Model governance: audit rights, explainability requirements, bias mitigation obligations
• Operational SLAs: uptime expectations, incident response times, support escalation
• Exit & transition: data return or deletion obligations, portability of data and models, continuity of service during transition

Technical checklist (short)

Quick technical checklist to include in an NDA'd data room review.

• Raw dataset descriptions and sample records
• Model architecture summary and training logs
• Evaluation metrics, confounding variable analysis, bias tests
• Reproducible training scripts or containerised artifacts

ML Engineer — role spec (Kenya-adapted)

Responsibilities: productionising ML models, data pipeline development, model monitoring and deployment on constrained infra (edge or low-cloud). Must-haves: strong Python/SQL, model deployment experience, familiarity with containerisation. Nice-to-haves: experience with satellite or sensor data, local-language NLP. Local considerations: hybrid Nairobi presence often required for cross-team collaboration.

• Key responsibilities: end-to-end model deployment, CI/CD for ML, observability
• Hiring signals: shipped production models, reproducible experiments, infra pragmatism
• Compensation note: stage-dependent and region-adjusted — discuss total package including equity

Data Scientist — role spec (Kenya‑adapted)

Responsibilities: experimental design, causal analysis, product-metric definition, lightweight model development and stakeholder communication. Must-haves: strong statistics, experimentation experience, storytelling with data. Local considerations: ability to work with noisy or sparse datasets and to support field-based pilots.

• Key responsibilities: analysis pipelines, KPI design, lightweight modelling
• Hiring signals: field pilot experience, ability to translate models into operational metrics

ML Product Manager — role spec (Kenya‑adapted)

Responsibilities: translate sector problems into model scope, prioritise features for low-bandwidth users, define success metrics and manage pilot rollouts with partners. Must-haves: product experience in fintech/agritech/healthtech, stakeholder management and procurement familiarity.

• Key responsibilities: product roadmaps, pilot scoping, partner coordination
• Hiring signals: prior pilot ownership, procurement cycle experience

10 interview prompts (local data engineering & product thinking)

Ten prompts to evaluate candidates’ ability to handle Kenyan datasets and product constraints.

• 1. Describe how you would build a credit-score model with sparse formal credit history — what alternative signals would you test and why?
• 2. How would you detect and mitigate bias in an agricultural yield model that underperforms for smallholder farms?
• 3. Explain a reproducible training workflow you would use when compute budgets are limited.
• 4. Walk through how you’d instrument model-monitoring for a deployed clinical triage model.
• 5. Design a minimal data schema to capture farmer input, harvest and buyer transactions for traceability.
• 6. How would you validate speech-to-text performance for Kiswahili in noisy mobile networks?
• 7. Give an example of a pragmatic offline-first UX decision for a low-bandwidth field-worker app.
• 8. Explain how you’d measure commercial impact for a demand-forecasting model used by FMCG agents.
• 9. How would you structure an A/B test for a lending decision model while ensuring fair treatment?
• 10. Describe a failure you encountered in production and the operational remediation steps you took.

A) Corporate innovation lead → startup

Subject: Pilot opportunity with [CorporateName] — interest in [startup name] Hi [Founder Name], I’m [Your Name], lead for digital innovation at [CorporateName]. We’re exploring a 3-month pilot on [use case: e.g., demand forecasting/clinical triage] and your work on [product area] looks relevant. Would you be open to a 30-minute call next week to discuss data needs and a pragmatic PoC scope? If helpful, I can share a template PoC scope and contracting checklist. Best, [Your Name] Follow-up 1 (5 days): Quick follow-up — any availability for a 30-minute call? Follow-up 2 (7 days later): If now’s not the right time, can you point me to a colleague or share a 1-page product brief?

B) Investor → founder screening intro

Subject: Intro from [InvestorName] — quick screening call? Hi [Founder Name], I’m [Your Name] with [InvestorName]. We invest in early-stage Kenyan AI/tech startups and would like a 20-minute intro to learn about product traction, data sources and the team. Could you share a one-pager and 30-minute slot this week? Key areas we’re curious about: dataset access, reproducibility, and who the early customers are. Regards, [Your Name] Follow-up 1 (3 days): Checking in on availability for a short call. Follow-up 2 (one week): If interested, please share a one-pager and a preferred time slot.

C) Government procurement office → inviting pilot proposals

Subject: Request for pilot proposals — citizen-service automation Dear [Official Name], [Agency] is seeking proposals for 6–12 week pilots to improve citizen-service intake and routing. Please submit a 2-page proposal describing scope, data handling, KPIs and an anonymisation approach by [date]. We prioritize solutions that demonstrate local-language support and low-bandwidth operation. Thank you, [Procurement Lead] Follow-up 1 (7 days): Reminder: proposal window closes in one week. Follow-up 2 (3 days before deadline): Final reminder and clarification on submission format.