Solomon Kimunyu

Experience

• 

  Kaggle

  Jan 2018 - Mar 2020

  Data Scientist

  1. Text classification: Built a model using natural language processing (SpaCy, NLTK & Fastai) toidentify potential toxic comments on Wikipedia, contributing to a safer online environment.2. Real-time Starfish Object Detection: Developed an object detection model (YOLOv5) for real-time starfish identification, enhancing research or conservation efforts.3. Health Recommender System: Built a health recommender system leveraging various similarity measures (Person, Jaccard, Cosine) to personalize health recommendations.4. Chronic Disease Analysis and Scoring: Developed a system for chronic disease scoring and dataanalysis using Python libraries (Pandas, Matplotlib, Seaborn, Plotly).5. Investment Return Rate Forecasting: Utilized machine learning (Fastai & Keras) to create a modelpredicting investment returns, aiding in informed financial decisions. Show less

• 

  Zindi

  Mar 2020 - Dec 2021

  Data Scientist

  1. Social media analysis: Leveraging NLP to analyze sentiment in Swahili conversations for brand insights.2. Environmental conservation: Developing AI models for automated turtle recognition and marine invertebrate classification.3. Bridging the digital divide: Engineering Wolof speech recognition to empower non-literate individuals.4. Infrastructure development: Utilizing AI for road segment identification from satellite imagery.5. Financial inclusion: Predicting individuals most likely to benefit from banking services in East Africa.6. Retail optimization: Designing solutions for real-time product removal prediction, enhancing customer experience.7. Social good: Collaborating on building algorithms to classify tweets about gender-based violence and predict air quality for informed decision-making.8. Agriculture: Developing AI models for passion fruit disease classification and crop classification in South Africa using satellite data.9. Financial services: Predicting customer churn, solar customer payments, and vehicle insurance claims to optimize resource allocation and customer service. Show less

• 

  Predictive Analytics Lab

  Jan 2021 - Mar 2022

  Data science trainer

  1. Mentored 20+ data science students, fostering their development in building efficient machine learning models and achieving a 99% success rate on their projects.2. Organized a machine learning competition on Kaggle that required students to apply their data science skills to predict future sales prices using advanced regression techniques, fostering engagement and practical application of skills for data science students.3. Developed and delivered data science curriculum, equipping students with skills in essential areas like Python programming, data analysis with libraries like pandas and NumPy, machine learning with algorithms like decision trees and random forests, deep learning using frameworks like TensorFlow and PyTorch, computer vision with tools like OpenCV, and recommender systems using collaborative filtering techniques. Show less

• 

  AICE Africa

  Aug 2021 - Jun 2022

  Data Scientist

  1. Developed a customer chatbot (OpenAI GPT-3) providing accurate COVID-19 information, streamlining information access for users.2. Built an AI maturity assessment tool (Django, HTML/CSS/JS) for clients, enabling data-driven evaluation of their AI readiness.3. Analyzed financial data (Python, pandas, SQL) for clients, identifying trends and insights using visualization libraries (Seaborn, Plotly, Matplotlib).4. Cleaned and preprocessed financial data (pandas, scikit-learn) to prepare it for further analysis and modeling.5. Developed a credit scoring system for a fintech client (Django, Postman, MongoDB), facilitating efficient risk assessment and decision-making.6. Created and deployed a product recommendation engine (NetworkX, Streamlit), personalizing product suggestions for users and enhancing customer experience. Show less

• 

  Equity Bank Limited

  Jun 2022 - Mar 2024

  Data Scientist

  1. Built a customer churn prediction model using Python (XGBoost, LightGBM), increasing retention by 55%, boosting deposits by 20%, and improving loan uptake by 5%, leading to a 50% reduction in churn.2. Developed a customer acquisition model that identified 15M potential customers, increasing customer acquisition by 50% and deposits by 25%.3. Developed a fraud detection model using Python (CatBoost, XGBoost, Pandas), preventing 11K fraudulent transactions, significantly strengthening security measures.4. Built a credit monitoring dashboard in Power BI and SQL, cutting default rates by 50% through proactive risk management.5. Architected an OCR-based document processing system (OpenCV, TensorFlow, PyTorch), reducing KYC verification time by 40%, cutting manual errors by 80%, and decreasing fraudulent applications by 35%.6. Created a 360-degree customer insights dashboard using Power BI and SQL, cutting branch service time by 50% and improving customer experience.7. Built a customer reactivation model (XGBoost, Random Forest), reactivating 1M dormant customers and reducing dormancy rates by 12%.8. Created a rule-based customer ranking model, onboarding 20K new customers monthly, driving sustainable growth.9. Analyzed 1B+ transactions using Python and SQL, driving a 50% increase in card spending and a 20% rise in new card users, while reducing transaction failure rates from 5% to 0.45%.10. Created an ecosystem onboarding model, ensuring customer funds remained within the financial ecosystem by optimizing supplier and distributor engagement.11. Designed an insurance cross-sell model, increasing insurance uptake by 50% and identifying 42K health insurance and 100K vehicle insurance prospects.12. Developed a treasury analytics report, increasing monthly forex customers from 50K to 90K and boosting deal value by 50%. Show less

• 

  Equity Bank Limited

  May 2024 - now

  Senior Data Scientist

  1. Developed credit scoring models across 5 countries, scoring 25M customers by integrating internal and external data, increasing loan disbursements by 59%, and achieving a 98% repayment rate while eliminating reliance on third-party scoring services.2. Designed a loan recovery prioritization model leveraging Gradient Boosting, increasing monthly recoveries through targeted collection strategies such as automated loan recovery processes using Python and SMS notification systems, increasing repayment rates from 94% to 98%, and reducing written-off loans by 40%.3. Developed an AI model for credit scoring, resulting in a 50% growth in lending activity and annual cost savings amounting to millions of dollars.4. Created fraud detection models surpassing industry standards, achieving a false positive rate below 20% and a true positive rate exceeding 80%.5. Engineered customer acquisition models that identified 13 million new prospects.6. Devised cross-selling models spanning banking, investment, and insurance products, potentially doubling organizational product revenues.7. Designed multiple dashboards supporting rapid decision-making across Finance, Customer Experience, Credit, Operations, and Insurance domains.8. Developed and deployed chatbots powered by Large Language Models (LLMs) integrated with Neo4j graph databases, enhancing data retrieval and customer interaction.9. Implemented Retrieval-Augmented Generation (RAG) to dynamically fetch relevant information, boosting response accuracy and improving customer satisfaction by 40% while reducing response times by 60%.10. Built a knowledge graph with Neo4j to organize customer support data, enabling efficient automated query resolution through LLMs.11. Fine-tuned LLMs (LLAMA and Deepseek) using LoRA (Low-Rank Adaptation) to create domain-specific chatbots for banking and finance, reducing customer support tickets by 30% and increasing customer engagement by 25%. Show less