Ronnie Ochieng
Data Scientist & Software Developer. Biomedical Engineer.
I am driven by the convergence of biomedical engineering, data science, and software development, passionately leveraging cutting-edge technologies to solve complex challenges. With a solid foundation in biomedical engineering and extensive experience as a data scientist and seasoned software developer, I have a proven track record of delivering impactful solutions across various domains.
Featured Projects
View selected projects below:
Automated Claims Processing Application
I developed an automated claims processing application built using Streamlit, a Python UI library. The application leverages `unstructured.io` library to extract information from structured and unstructured documents i.e. excel workbook, and PDF documents. Google's Gemini 1.5 Pro large language model is used to extract information in a structured format for downstream tasks. Business logic is implemented to identify duplicate claims, fraudulent claims, and suspicious claims. The system takes a treaty document, borderaux, and treaty slip document as input and processes the claims to determine whether they are fraudulent or legitimate, ultimately acception or rejecting the claims based on the results.
View Application
AI Product Sizing & Quotation Web Application
I collaborated on developing an AI-powered web application for sizing and quoting Davis & Shirtliff solar power backup products, such as Dayliff inverters and solar panels. The application utilizes an AI agent built with LangGraph, powered by Llama3, a large language model that reasons over web information to deduce the average power consumption of items. It integrates a retrieval-augmented generation (RAG) system powered by Llama3, with embeddings stored in a FAISS vector database, and uses a BM25 and FAISS ensemble retriever. The system generates structured output using Pydantic models, ensuring consistency in providing power backup solution quotations.
View Web Application
SomeshaAI ECDE AI-Powered Content Generation Web Application
Collaborated on developing a content generation system for the Early Childhood Development Education (ECDE) sector in Kenya. The system takes in a topic i.e. Numbers and generates a story, which is then coupled with background music and images representing the various scenes in the story. The final output is a video that can be used to teach children in the ECDE sector.
View Web Application
MamaPesa Chatbot
Developed MamaPesa Chatbot system using Llama3, that provides business growth advice to women entrepreneurs, facilitating asset financing inquiries, and autonomously researching and initiating purchases from vetted vendors based on user credit scores, enhancing the application capability to support small business development.
View HuggingFace Model
Dyula-French Neural Machine Translation System
Developed a neural machine translation system for the low-resource Dyula language to French, utilizing TensorFlow-Keras framework. Implemented a custom transformer architecture comprising 6 layers and 8 attention heads in both encoder and decoder components.
View GitHub Repository
Optimized Retrieval System using Hybrid search
I developed an optimized retrieval system using hybrid search that combines BM25 and FAISS retrievers generating both dense and sparse vector embeddings to ensure relevant documents are returned upon user query.
View GitHub Repository
Automated Number Plate Recognition System
I developed an automated number plate recognition system that is capable of detecting and recognizing number plates in images using OpenCV's Canny edge detection algorithm and Tesseract OCR. The system is designed to automate Kenyan vehicles' license plates recognition to automate parking fee payment systems.
View Streamlit Application
AgriXpert Bot
Streamlit application that shows a proof of concept how large language models can be used to provide remedies to farmers' queries. The application supports Swahili, English, and French languages. The model was fine-tuned on an agricultural dataset. The chatbot system has been used in building AgriXpert Bot system which is a voice-SMS chatbot that provides agricultural information to farmers in Kenya.
View Streamlit Application
Dialogue Summarization and Detoxification System
Developed a dialogue summarization model that generated less toxic content with Meta AI's hate speech reward model. The reward model is a binary classifier that predicts either "not hate" or "hate" for the given text. It uses Proximal Policy Optimization (PPO) to fine-tune and reduce Flan-T5 model's toxicity.
View GitHub Repository
Swahili Speech Recognition System
Led a team of three individuals to develop a robust and efficient Swahili Speech Recognition system that can be applied across various domains, including transcription services, voice assistants, language learning applications, and more. The system was built to enhance acessibility to technology for Swahili speakers, enabling seamless interaction with digital devices through spoken language.
View Hugging Face Repository
ClassifyMe Backend APIs
Architected and implemented the backend infrastructure for ClassifyMe, a cross-platform application offering personalized course, university, and career recommendations to high school graduates. Designed and deployed RESTful APIs with comprehensive documentation, facilitating seamless front-end development team integration and ensuring scalable, efficient data processing and recommendation delivery.
View Browsable APIs
E-waste Detection System
I designed a system to facilitate the responsible disposal of e-waste. Users will be able to contribute their e-waste and, in return, receive points equivalent to the assessed value of the items, determined through cutting-edge computer vision techniques.
View Streamlit Application
AfyaMumBot - Personalized Pregnancy Support Chatbot
Developed AfyaMumBot, a Whatsapp-based AI-powered chatbot system designed to offer personalized information and support to expectant mothers throughout their pregnancy journey. It leverages a large language model; Llama2 to provide accurate and helpful responses to user inquiries related to obstetrics and gynecology.
View GitHub
Yatsa Data Analytics
YatsaApp is a car tracking solution that provides real-time GPS tracking and monitoring services for vehicles. The project's main aim was to provide valuable insights derived from the data gathered from the Yatsa fleet of vehicles. The insights were used to improve the company's operations and customer service.
View Analytics Dashboard
BebaPeleka
BebaPeleka project is a transformative solution aimed at tackling the issue of inadequate compensation that farmers often face when selling their produce through intermediaries. The primary objective of this project was to empower farmers by enabling them to monitor and receive fair remuneration for their agricultural efforts. I led a team of four individuals including myself in modelling data gathered from farmers' produce receipts. We developed an entity relationship diagram (ERD) and a relational database schema for the BebaPeleka database. I designed a Figma prototype for the BebaPeleka mobile application.
View Figma PrototypeWork Experience
Freelance Machine Learning & Backend Engineer
Upwork
February 2024 - Present
I freelance on Upwork as a Machine Learning & Backend Engineer. I have worked on various projects including developing machine learning models, building web applications, and creating APIs. I have experience working with a client for a project that involved developing agentic workflows using CrewAI.
Biomedical Engineering Intern
Mater Misericordiae Hospital
Internship
January 2024 - March 2024
As a Biomedical Engineering Intern at Mater Misericordiae, I was responsible for repairing and maintaining various medical equipment such as baby incubators, dialysis machines, ultrasound, suction machines, and ventilators. This involved using my knowledge of engineering principles and medical technology to diagnose problems with the equipment, perform necessary repairs, and ensure that the equipment was operating at optimal levels.
Data Analyst
Tingle Software: Yatsa
March 2023 - April 2023
As a data analyst at Tingle Software, I was responsible for analyzing data from the Yatsa fleet of vehicles, providing valuable insights that were used to improve the company's operations and customer service.
Biomedical Engineering Intern
Kenyatta National Hospital
Attachment
January 2023 - March 2023
As a Biomedical Engineering Intern at Kenyatta National Hospital, I was responsible for repairing and maintaining various medical equipment such as baby incubators, dialysis machines, ultrasound, suction machines, and ventilators. This involved using my knowledge of engineering principles and medical technology to diagnose problems with the equipment, perform necessary repairs, and ensure that the equipment was operating at optimal levels.
Education
Kenyatta University.
Biomedical Engineering
5th Year Biomedical Engineering Student
KamiLimu
KamiLimu Mentee
KamiLimu is a non-profit organization whose mission is to upskill tertiary-level tech students with competencies in personal and professional development, industry-relevant ICT skills, innovation, scholarship, and community engagement, in an effort to nurture employability, enhance student learning outcomes, and increase global competitiveness.
DeepLearning.AI
Jan 2024
Completed AI for Medical Diagnosis course offered by DeepLearning.AI
Coursera
Sep 2023
Completed Image Super Resolution using Autoencoders in Keras course offered by Coursera Project Network.
Coursera
July 2023
Completed Unsupervised learning, recommender systems, and reinforcement learning course in the Machine Learning Specialization course offered by DeepLearning.AI
October 2022
Completed Supervised Machine Learning: regression and classification Course in the machine learning specialization course offered by DeepLearning.AI
Goethe Institut
I learnt Deutsch language right from the beginners level A1 to A2. I have an elementary understanding of the German language and culture.