Cleavestone Adungo

Passionate and results-driven Data Analyst & AI Practitioner with over 2 years of hands-on experience, actively working on and documenting real-world projects. My work spans from data analysis using Python, SQL, and Power BI, to predictive modeling, and advanced LLM fine-tuning and prompt engineering.

I also regularly participate in data science competitions on platforms like Kaggle and Zindi, applying my skills to real-world problems and sharpening my edge.

A committed and curious learner, I’m constantly exploring new tools and staying up to date with trends in data, AI, and machine learning, ensuring I deliver value through innovation and practical insight.

• Certificate in MLOPS from Udemy
• Certificate in LLM Engineering from Udemy
• Certificate in Foundational Generative AI from Ineuron

• Movie recommender app - a semantic movie recommendation system powered by LanceDB, Sentence Transformers, and Streamlit This app suggests movies based on the similarity of user queries to movie descriptions using vector embeddings.
• Amini Soil Challenge hosted on the Zindi platform, ranked 40th out of 400 participants in a machine learning competition focused on predicting soil nutrient availability and crop yield optimization.
• ETL Pipeline : This project builds a reliable ETL pipeline using Apache Airflow and Python. It extracts real-time weather data from the OpenWeatherMap API, transforms it into a structured format, and loads it into MongoDB Atlas. The setup enables efficient storage and future analysis of weather trends in the cloud.
• Data Analysis (SQL) - Built an interactive SQL practice project with realistic datasets to strengthen querying skills. It features three tables: customers, products, and orders The project covers essential SQL topics like JOINs, GROUP BY, subqueries, and window functions through practical exercises.
• Medical Chatbot - This medical chatbot leverages Retrieval-Augmented Generation (RAG) to provide accurate, context-aware responses. It uses LangGraph to manage conversational flow, integrates Pinecone for vector-based document retrieval, and runs LLaMA models via Groq for fast, high-quality generation