Hamburg, Germany
ML Engineer and GenAI practitioner with hands-on experience building production-grade AI systems and full-stack web applications. Backed by 4 peer-reviewed publications and practical experience across AWS, Azure, and modern frontend frameworks.
Selected Work
Client Project
Full-stack food delivery platform built for a food service business. Features user authentication, a full menu system, custom order requests, and real-time order tracking from kitchen to door.
The client needed a way to take orders online and give customers visibility into their delivery status. I built the entire platform from scratch using Next.js and TypeScript — auth with protected routes, a menu with categories and filtering, a custom order flow for special requests, and a live status tracker. Deployed on Vercel with CI/CD on every push.
Client Project
Professional multi-page website for a Ghanaian publishing house. Covers service listings, author submission flow, blog, and contact.
Built for a publishing house in Ghana that needed a credible online presence to attract authors. I translated their business requirements into a clean Next.js site with a service showcase, a step-by-step publish-with-us flow, blog, and newsletter subscription. Source code is private per client agreement.
Client Project
Elegant event planning platform with booking flow, service showcase, portfolio gallery, and client testimonials.
An event planning company in Ghana needed a site that felt as premium as their service. I focused heavily on design — large imagery, smooth scroll, a services breakdown, a portfolio gallery of past events, and a booking form. Built with Next.js and TypeScript, deployed on Vercel.
AI / ML
Production RAG pipeline over 389 arXiv papers — 1,500 semantic chunks in ChromaDB, 0% zero-result rate, deployed with MLflow tracking and CI/CD to Hugging Face Spaces.
Built a full retrieval-augmented generation system to answer technical questions over a corpus of ML research papers. Ingested and chunked 389 arXiv PDFs into ChromaDB with sentence-transformer embeddings. Built a FastAPI backend with dual-mode LLM (Mistral-7B local / API fallback), MLflow tracking across 3 evaluation runs, and automated CI/CD deployment to Hugging Face Spaces.
AI / ML
5-agent LangGraph StateGraph with typed state, conditional routing, and a Critic retry loop. 85% quality score at 7.6s end-to-end.
Designed a multi-agent system for cybersecurity research using LangGraph. The Supervisor routes queries to specialised agents — RAG (ChromaDB over security papers), WebSearch (Tavily), CodeAnalyst, and a Critic that retries low-quality responses. Built with Groq LLaMA 3.1 for speed, FastAPI serving, MLflow per-query tracking, and GitHub Actions CI/CD to Hugging Face Spaces.
Client Project
Modern, user-friendly site for Amdor Lodge, highlighting lodging options, charm, and booking essentials. Focused on clean UI/UX with dynamic elements to drive inquiries and position the business as a welcoming Kasoa destination.
Developed a website for Amdor Lodge, a hospitality platform showcasing lodging services, amenities, and booking features. Leveraged modern web technologies including HTML, CSS, JavaScript, and responsive design frameworks to create an intuitive user interface optimized for desktop and mobile devices. Integrated dynamic elements like image galleries and contact forms to enhance user engagement and streamline inquiries, demonstrating proficiency in front-end development, UI/UX principles, and deployment best practices for real-world client projects.
Data Visualization
Live weather sensor analytics dashboard — Open-Meteo Archive API, PostgreSQL, 10 analytical SQL queries, Streamlit + Plotly, CSV/Excel export. 6 European cities, 330 readings/refresh.
Built an interactive real-time sensor data analytics dashboard using Streamlit, showcasing machine learning and IoT data visualization skills. Features live data feeds, interactive charts, and key performance metrics for monitoring sensor streams, ideal for IoT applications and predictive maintenance.
AI / ML
EfficientNet-B3 vision classifier for industrial casting defect detection — 99.7% accuracy, 100% recall on 715 test images, LIME explainability, Azure ML deployment pipeline, MLflow tracking.
Developed a high-accuracy image classification model for industrial quality inspection using EfficientNet-B3. Achieved 99.7% accuracy and 100% recall on a test set of 715 images, ensuring reliable defect detection. Implemented LIME for model explainability, providing insights into predictions. Deployed the model using Azure ML with a robust CI/CD pipeline and MLflow tracking for performance monitoring and iterative improvements.
AI / ML
LSTM Autoencoder for industrial sensor anomaly detection on NASA CMAPSS — AWS SageMaker endpoint, live Grafana + Prometheus monitoring, MLflow tracking, SHAP explainability, SageMaker Pipelines DAG.
Built an LSTM Autoencoder for anomaly detection on the NASA CMAPSS dataset, achieving high accuracy in identifying anomalies in industrial sensor data. Deployed the model as an AWS SageMaker endpoint with a CI/CD pipeline using SageMaker Pipelines. Implemented live monitoring with Grafana and Prometheus, and used SHAP for model explainability to provide insights into feature importance and model predictions.
AI / ML
XGBoost RUL forecasting for turbofan engine predictive maintenance — NASA CMAPSS, SHAP explainability, AWS SageMaker deployment, Grafana + Prometheus monitoring
Developed an XGBoost model for Remaining Useful Life (RUL) forecasting in predictive maintenance using the NASA CMAPSS dataset. Achieved high accuracy in predicting the remaining useful life of turbofan engines, enabling proactive maintenance scheduling. Deployed the model on AWS SageMaker with a CI/CD pipeline, and implemented live monitoring using Grafana and Prometheus. Utilized SHAP for model explainability, providing insights into feature importance and model predictions to enhance trust and interpretability.
AI / ML
LangChain-style agent for AI-powered engineering design — arXiv + Semantic Scholar knowledge base, ChromaDB, 4-tool pipeline (parse → retrieve → generate → evaluate), PDF + Word export, MLflow tracking, CI/CD via GitHub Actions, live on HF Spaces.
Built a generative design assistant for engineering applications using a LangChain-style agent architecture. The system ingests and processes technical papers from arXiv and Semantic Scholar, storing them in ChromaDB for retrieval. The agent pipeline includes parsing user queries, retrieving relevant information, generating design suggestions, and evaluating outputs. Features PDF and Word export capabilities, MLflow tracking for performance monitoring, and CI/CD deployment via GitHub Actions to Hugging Face Spaces.
Technologies
Academic Background
Research
Talks & Presentations
Get in Touch