Volunteer AL/ML Engineer

AI/ML Engineer (Volunteer)
Department: Technology / Data Science
Reports To: Technology Lead /CPO
Location: Remote
Commitment: 10 to 20 hours per week
Type: Volunteer
About Black Mental Health Matters Inc
Black Mental Health Matters Inc is a nonprofit organization dedicated to advancing mental health awareness, resources, and support within Black communities. We leverage technology, including AI-driven tools, to expand access to mental health information and improve outcomes for the communities we serve.
Role Overview
We are seeking a motivated AI/ML Engineer to join our technology team as a volunteer. In this role, you will evaluate and enhance existing machine learning processes, perform statistical analysis to resolve data set challenges, and improve the accuracy and predictive capabilities of our AI-driven software. This is an opportunity to apply your technical expertise toward a mission-driven cause that directly impacts mental health outcomes in underserved communities.
Key Responsibilities
• Consult with managers to determine and refine machine learning objectives.
• Design machine learning systems and self-running AI software to automate predictive models.
• Transform data science prototypes and apply appropriate ML algorithms and tools.
• Ensure that algorithms generate accurate user recommendations.
• Convert unstructured data into useful information through auto-tagging of images and text-to-speech conversion.
• Apply a clear understanding of Machine Learning Operations (MLOps).
• Solve complex problems involving multi-layered data sets, and optimize existing ML libraries and frameworks.
• Apply knowledge of transformer models where relevant.
• Support model integration with UI and manage model deployment.
• Apply solid understanding of data structures.
• Build and automate local data ingestion pipelines to extract, clean, and compress audio/video assets from media archives.
• Implement and optimize open-source Automatic Speech Recognition (ASR) and speaker diarization tools (e.g., Whisper.cpp, Faster-Whisper, PyAnnote) to generate accurate transcripts from multi-speaker recordings.
• Design scripts to clean, filter, and structure raw transcripts into instruction-tuning datasets (JSON/JSONL) for conversational AI training.
• Execute parameter-efficient fine-tuning (PEFT), including QLoRA, on open-source models (e.g., Llama 3 8B) using frameworks such as Unsloth, Axolotl, or Hugging Face TRL.
• Manage and optimize compute/memory usage to run training and evaluation within the constraints of commodity hardware (e.g., single RTX 3090/4090 or Apple Silicon).
• Maintain data sovereignty by ensuring training data and model weights remain localized, secure, and air-gapped from commercial cloud APIs.
• Develop ML algorithms to analyze large volumes of historical data to generate predictions.
• Run tests, perform statistical analysis, and interpret results.
• Document machine learning processes clearly for internal knowledge-sharing.
• Stay current with developments and best practices in machine learning.
Qualifications & Skills
• Bachelor's degree (or equivalent experience) in Computer Science, Data Science, Machine Learning, or a related field.
• Demonstrated experience building and deploying machine learning models.
• Proficiency in Python and common ML frameworks (e.g., TensorFlow, PyTorch, scikit-learn).
• Familiarity with transformer models and modern NLP techniques.
• Experience building, fine-tuning, or working with Large Language Models (LLMs).
• Experience with model deployment and integration with user-facing applications.
• Hands-on experience with the Hugging Face ecosystem (Transformers, PEFT, TRL, BitsAndBytes).
• Experience running, quantizing (GGUF/AWQ), and fine-tuning models locally; familiarity with memory-saving libraries like Unsloth is a plus.
• Experience with audio processing tools (e.g., FFmpeg, yt-dlp) and multi-speaker dialogue formatting.
• A scrappy, resourceful mindset — comfortable optimizing open-source tools under resource constraints rather than relying on costly cloud infrastructure.
• Strong understanding of data structures and algorithms.
• Ability to work independently in a remote, volunteer-based environment.
• Strong communication skills and comfort documenting technical processes.
• Genuine interest in mental health advocacy and community impact work.
Nice-to-Haves
• Experience deploying local inference servers (e.g., Ollama, vLLM, LM Studio) for testing and evaluation.
• Familiarity with containerization (Docker) for reproducible local environments.
• Knowledge of evaluation frameworks for specialized domain LLMs to help prevent hallucination.
What You'll Gain
• Meaningful, mission-driven volunteer experience applying AI/ML skills for social good.
• Flexible remote schedule (10–40 hrs/week).
• Collaboration with a passionate, cross-functional team.
• Opportunity to contribute to real-world AI applications supporting mental health access.
How to Apply
Interested candidates should submit their application and resume to hr@bmhm.org.

Work arrangement
No