Medical Doctor • Applied AI/ML Engineer • AI Healthcare Research

I build safety-aligned AI systems for clinical deployment, specializing in multimodal screening agents (voice/text), fine-tuned LLMs, and RAG architectures for mental health and substance use disorder interventions in resource-constrained settings.

Frontier AI Safety & Clinical Deployment
Engineering robust, interpretable AI systems at the intersection of large language models and healthcare. My work focuses on:

• Hallucination mitigation & bias reduction in clinical LLMs
• Multimodal screening agents (voice + text) for mental health assessment
• RAG architectures for evidence-grounded clinical decision support
• Safety alignment for real-world healthcare delivery in low-resource contexts

End-to-end ML pipeline predicting 30-day hospital readmission using Random Forest + SMOTE. Built with scikit-learn and deployed as a Streamlit web app.

• Tech: Python, pandas, scikit-learn, imbalanced-learn, Streamlit
• Impact: 89% precision for high-risk flagging with class imbalance handling

Real-time clinical RAG system for evidence-grounded medical decision support

• ⚡ Latency: <200 ms query response time
• 📊 Hallucination Reduction: ~40% reduction in GPT-4 hallucinations through evidence grounding
• 🎯 Accuracy: Evidence-backed answers with confidence scoring

Safety-aligned RAG chatbot for youth with substance use disorders, incorporating culturally relevant language and escalation protocols.

• Hallucination Reduction: ~40% reduction in irrelevant responses
• Safety: Multi-stage validation with human clinician escalation pathways

Core AI/ML
Safety-aligned LLMs • Fine-tuning (PEFT, LoRA) • RAG architectures • Multimodal models • Hallucination mitigation • Bias detection & mitigation • Model interpretability (SHAP, LIME)

ML Infrastructure
Python (PyTorch, TensorFlow, Pandas, Scikit-learn) • LangChain • Pinecone/Vector DBs • REST APIs • Model deployment pipelines

Clinical AI Applications
Mental health assessment • Safety benchmarking • Escalation protocol design • Low-resource NLP • Clinical trial design

Tools & Platforms
GitHub • Labelbox • Cloud compute • Jupyter/Colab • Streamlit

• Olufadewa II, Adesina MA, Olawoyin MT, Usman UB, et al. "Developing the GENSCORE Unimodal Voice Feature Model: Automated Depression Screening in Multimodal Framework." arXiv:7031573 (2025).
• Olufadewa II, Adesina MA, Oladejo EA, Usman UB, et al. "Finetuning Large Language Models for Automated Depression Screening in Nigerian Pidgin English: GENSCORE Pilot Study." arXiv:7032000 (2025).

• M.B.B.S. (Bachelor of Medicine, Bachelor of Surgery) — University of Ibadan, Nigeria | 2016–2023
• AI in Healthcare Specialization — Stanford University School of Medicine Online | 2025
• Machine Learning Specialization — DeepLearning.AI | 2025
• AI Engineering Professional Certificate — IBM | 2025

Associate Research Scientist | Applied AI/ML Engineer
Slum and Rural Health Initiative (SRHIN), Nigeria | 2025–Present

• Co-authoring research on fine-tuning LLMs for Public health Application
• Architecting AI-assisted screening workflows with escalation pathways to human clinicians
• Developing AI-driven intervention frameworks for funded healthcare delivery research
• Co-developing translational protocols to assess clinical safety and utility of digital health tools

AI/ML Engineer — Independent Developer & Consultant | 2024–Present
ML engineering, predictive modeling with SHAP interpretability, and generative AI/RAG pipeline deployment for clinical applications.

• AI Safety & Alignment in clinical settings
• Low-resource NLP for African languages and dialects
• Multimodal AI for mental health screening (voice, text, behavioral signals)
• Explainable AI (XAI) for clinical decision support
• Digital health equity in resource-constrained environments

• Federal Scholarship Board Award (2020)
• NNPC/Chevron National University Scholarship (2017)
• Shell Petroleum Development Company Scholarship (2017)

• Email: buthman98@gmail.com
• LinkedIn: linkedin.com/in/uthman-babatunde-m-d-126582286
• GitHub: github.com/uthy4r

Building trustworthy AI systems for healthcare delivery in Africa and beyond.