About Me
Hello! I'm Mahdi Khemakhem, a Machine Learning Engineer and developer who builds scalable, AI-powered systems across diverse domains. My work spans healthcare, research, finance, and blockchain technologies, where I develop solutions that combine technical innovation with practical impact.
At the Precision Brain Health Initiative, I develop a range of systems—from information retrieval architectures to natural language processing pipelines. Beyond my current role, I've created trading systems for blockchain networks, built data platforms for large-scale research initiatives, and implemented automated workflows that significantly enhance productivity.
With a Master's in Computer Science specializing in Data-Centric Computing and a Bachelor's in Neuroscience, I bring a multidisciplinary approach to problem-solving. I'm proficient with Python, ML/DL frameworks, cloud architecture, containerization technologies, and more, allowing me to build end-to-end solutions that deliver concrete value in diverse environments.
Education
My academic background and qualifications.
Master of Science in Computer Science
Boston University
Specialization in Data-Centric Computing, 3.82/4.0 GPA
- Coursework: Algorithms, Networks, Databases, Machine Learning, Deep Learning, Object-Oriented Programming, Neural Modeling, Natural Language Processing
Bachelor of Arts in Neuroscience with Honors
Boston University
Graduated Summa Cum Laude with a 3.95/4.0 GPA
- Presidential Scholarship recipient
- Dean's List all semesters
Work Experience
My professional journey and roles I've held over the years.
Machine Learning Support Engineer
Precision Brain Health Initiative, Boston University Medical School
Applying machine learning techniques to develop solutions in the dementia research space.
- Academic Chatbot: Architected a retrieval augmented-generation (RAG) system, to enrich LLM prompts with real scientific text, and support responses with accurate citations, improving answers to research related questions
- Implemented a pipeline to crawl, and embed 3M+ articles using transformers and Pinecone vector database.
- Deployed a Gradio interface using LLMs (OpenAI, Ollama, vLLM) to support internal research queries.
- Literature Review Crawler: Developed a Python-based web crawler using LLMs, to automate paper discovery and selection, scaled to process thousands of papers daily, supporting grant-writing efficiency for researchers.
- Docker Microservices: Developed natural language (NLP) and audio processing containers that provide abstracted REST API endpoints to popular libraries (SpaCy, NLTK, SpeechBrain) reducing workflow integration time.
Research Developer
Davos Alzheimer's Collaborative
Developed systems to support research cohorts, and team operations.
- Realtime Reports: Delievered realtime project management reports to leadership, leveraged Monday.com GraphQL API, Webhooks, Azure Functions, and PowerBI push datasets to maintain an up-to-date dashboard.
- Task Classification: Implemented a realtime SVM classifier, triggered by webhooks and powered by Azure Functions, to label hundreds of tasks daily, increasing labeling compliance and improving data completeness.
- Data Delivery: Automated participant data upload, using AWS Lambda, S3 and EventBridge.
Research Assistant
Kolachalama Lab, Boston University
Focused on LLM development for neuroscience research.
- NeuroLLM: Fine-tuned a Falcon-40B LoRA, improving base performance by 50% on PubMedQA benchmark.
- Led a team of three engineers to develop an internal site hosting LLMs fine-tuned on scientific neuroscience text.
Skills & Technologies
My technical skill set and expertise across different domains.