About Me
I build AI that works in the real world. As a Senior AI Systems Engineer with 5+ years of technical execution, I specialize in the full lifecycle of Machine Learning—from production-grade multi-agent orchestration to robust LLMOps deployment on AWS.
While my recent work has revolutionized how AI serves the agricultural and climate sectors, my core expertise lies in architecting self-service platforms and autonomous agents. I don't just build models; I build solutions that reduce consultation times, increase efficiency, and scale with the business.
My technical foundation in LLMOps, RAG architectures, Agentic AI, and AWS infrastructure (SageMaker, EC2, RDS) enables me to deliver measurable ROI across any high-stakes industry—FinTech, Logistics, SaaS, or beyond. Ranked #32 globally on Zindi, I bring a proven ability to translate complex ML systems into strategic business value.
Core Expertise
Technical Skills
AI & Generative Engineering
Data Science & Machine Learning
Cloud & Infrastructure (MLOps)
Backend & Development
Featured Projects
Building production-ready AI systems that solve real-world problems in Agritech and Climate-tech.
Rhea API
AI-Powered Agricultural Recommendation API for East African Farmers
Problem
Solution
Key Achievements
- •10K+ farmers served across Kenya's 47 counties
- •4 parallel AI recommenders (fertilizer, pH, organic, foliar)
- •55 tomato diseases supported in pathology analysis
- •Production deployment on AWS ECS with CI/CD
- •Multi-stage Docker optimization with uv
Tech Stack
Rhea - AI-Powered WhatsApp Advisory System
Multimodal Agronomic Assistant Serving 10K+ Farmers
Problem
Solution
Key Achievements
- •10K+ farmers served across Kenya
- •40% reduction in consultation time
- •Bilingual support (Swahili/English)
- •Real-time AI responses via WhatsApp
Tech Stack
Scan to Try
WhatsApp AI Chatbot
QwenMkulima - Fine-Tuned Agricultural LLM
Specialized LLM Bridging Global AI with Local Agricultural Context
Problem
General-purpose LLMs lack domain-specific knowledge for smallholder agriculture in East Africa, leading to generic and often inapplicable advice.
Solution
Fine-tuned Qwen2.5-1.5B using QLoRA on 6,000+ agricultural instruction samples to create a domain-specific LLM optimized for soil fertility, pest management, and climate-smart agriculture.
Key Achievements
- •4-bit NF4 quantization for efficient inference on T4 GPUs
- •Adapter size reduced to ~150MB
- •Improved domain recall and grounding for farmer advisory tasks
- •Self-correcting advisory system for complex agricultural decisions
Tech Stack
DOXA - PDF-to-Knowledge RAG Pipeline
Efficient RAG Foundations for Document Intelligence
Problem
Organizations struggle to extract actionable insights from large volumes of unstructured PDF documents.
Solution
Architected a Retrieval-Augmented Generation (RAG) system using LangChain and OpenAI to transform unstructured PDF data into a searchable knowledge base with conversational interface.
Key Achievements
- •FAISS vector search for low-latency retrieval
- •Modular Python components for scalability
- •Implemented Semantic Chunking and Hierarchical Parent-Child Retrieval to preserve deep contextual integrity across dense documents.
- •Integrated Hybrid Search (BM25 + Dense Embeddings) paired with a Cross-Encoder Reranking pipeline to maximize retrieval accuracy and minimize hallucination vectors.
Tech Stack
Zoloni - Enterprise Expense Automation via WhatsApp
Enterprise Expense Automation for Field Teams
Problem
Manual financial data entry created friction for field teams, leading to delayed reporting and data inconsistencies across enterprise operations.
Solution
Architected a high-concurrency financial tracking platform utilizing a robust, deterministic state-machine architecture via Flask and Twilio, backed by Supabase for resilient session state persistence.
Key Achievements
- •Engineered an asynchronous state-machine framework to manage complex, multi-turn user session states over WhatsApp webhooks flawlessly.
- •Enabled real-time financial data capture via WhatsApp messaging
- •High-concurrency data persistence with Supabase
- •Streamlined enterprise expense reporting workflows
Tech Stack
AWS Infrastructure Migration
Scaling Agritech AI from On-Prem to Cloud
Problem
Legacy on-premise systems at Rhea limited scalability and increased operational overhead.
Solution
Designed and executed migration of ML infrastructure to AWS, leveraging EC2 for compute, SageMaker for model training/deployment, and implementing MLOps best practices.
Key Achievements
- •Zero-downtime migration
- •Improved system reliability and scalability
- •Cost optimization through right-sizing
- •Established CI/CD pipelines with GitHub Actions
Tech Stack
Professional Experience
Consultant AI Engineer & Technical Lead (Freelance / GDSC)
Freelance / Google Developer Student Clubs
- •Architected end-to-end ML models, built automated data workflows, and led technical workshops on machine learning deployment for over 50+ junior developers via Google Developer Student Clubs.
Lead AI Engineer & Data Scientist
Rhea Soil Management
- •Spearheaded the end-to-end architecture and deployment of an intelligent, multimodal agricultural ecosystem serving 10K+ farmers via a containerized WhatsApp interface.
- •Built high-performance async microservices using FastAPI and Docker deployed on AWS (ECS/EC2), reducing manual consultation workflows by 40%.
- •Integrated advanced Retrieval-Augmented Generation (RAG) for localized agronomic intelligence alongside edge-optimized Computer Vision models for zero-latency crop disease diagnosis.
- •Designed robust MLOps pipelines with automated model monitoring, cloud resource optimization, and strict compliance with the Kenya Data Protection Act.
Computer Vision Engineer
Vunatec
- •Developed ML models for automated mango grading
- •Deployed models for real-time inference in production
- •Optimized for variable lighting and constrained hardware
- •Reduced post-harvest losses through AI automation
Top 100 Data Scientist
Zindi
- •Ranked #32 globally out of 100,000+ AI builders (Top 0.05% worldwide) on the Zindi competitive machine learning platform.
- •Competitive ML challenges on real-world datasets
- •Advanced feature engineering and ensemble modeling
Education & Achievements
Education
BSc Computer Science
Laikipia University
Kenya • 2019 - 2023
Certifications
Certified Professional Data Scientist
DataCamp
AI Engineer
DataCamp
Data Scientist with Python
DataCamp
Key Achievements
#32 Global Ranking on Zindi
Ranked #32 globally out of 100,000+ AI builders (Top 0.05% worldwide) on the Zindi competitive machine learning platform.
10K+ Farmers Served
Through AI-powered advisory systems
40% Efficiency Gain
In agricultural consultation workflows
Production AI Systems
Deployed at scale in Agritech
Let's Build Something Impactful Together
I'm always open to discussing new projects, opportunities, or collaborations in AI/ML and Agritech.