ML Ops / Infrastructure Fellow (Health)

About FairwAI:
Stealth mode mission driven. FairwAI is building the AI bias detection and compliance platform that industries will rely on to ensure fairness in hospitality, and healthcare. We operate at the intersection of AI, policy, and accountability — backed by a team from MIT, and Harvard. Our work blends rigorous research, thoughtful product development, and ethical innovation at global scale.

Team
Shaurjya Mandal – Head of Research, AI Systems & Clinical Innovation
Role: Shaurjya leads FairwAI's research function and the development of its AI methodology — including structured retrieval architectures, sentiment and bias detection systems, and validation frameworks for fairness in clinical and high-stakes AI environments. He ensures FairwAI's research meets the standards of peer-reviewed clinical AI while remaining accessible and useful in real-world deployment.

Bio: Shaurjya Mandal is a machine learning research scientist at Mass General Brigham and Harvard Medical School, where he develops and evaluates AI systems for clinical settings. His research sits at the intersection of responsible AI, healthcare equity, and applied machine learning — with published and applied work in algorithmic bias, fairness in clinical decision support, and AI safety for vulnerable populations. He holds an M.S. in Artificial Intelligence from Carnegie Mellon University, where he conducted research at the Robotics Institute on machine perception and learning systems. Prior to his work at Harvard, Shaurjya served as a research engineer at India's Defence Research and Development Organisation (DRDO) and contributed to AI policy and infrastructure projects across the MENA region, including the design of responsible AI deployment frameworks for healthcare and insurance systems in the United Arab Emirates. He is a Nucleate Pittsburgh fellow, a recognized convener in early-stage biotechnology and AI commercialization. At FairwAI, Shaurjya bridges frontier clinical AI research with practical, ethical deployment — ensuring that what we build for the most underserved populations is also held to the highest scientific standard

Mr. Dixon – Chief Strategist Role: Visionary founder of FairwAI, Mr. Dixon leads overall strategy, mission alignment, institutional partnerships, and compliance innovation. He ensures the platform remains focused on real-world fairness, risk mitigation, and public accountability—rather than abstract DEI. Bio: Mr. Dixon is a serial founder, public interest technologist, and global strategist known for building impact ventures at the intersection of technology, governance, and equity. He is currently a strategic advisor at MIT, where he collaborates across departments such as the MIT Media Lab and Martin Trust Center. Mr. Dixon has served as a policy strategist to President John Mahama of Ghana, worked with the U.S. Department of Energy on distributed energy infrastructure, and led multi-million-dollar public-private partnerships with organizations like GE Healthcare, USAID, and UNICEF. Mr. Dixon is the founder of SolarFi, a social enterprise that builds portable solar-powered infrastructure for clinics, schools, and underserved communities across Africa providing energy to over 600,000 people. Under his leadership, SolarFi was awarded major U.S. government contracts and featured by MIT and the UN. He has also advised on diplomatic and philanthropic partnerships.

FairwAI is Mr. Dixon’s most urgent venture yet. Inspired by lived experiences of racial bias and systemic discrimination in global travel, hiring, and healthcare systems, he founded FairwAI to reframe fairness as a compliance and safety issue, not a corporate DEI trend. Antonio brings a rare combination of field-tested grit, institutional access, tech, diplomacy, and global coalition to ensure accountability becomes enforceable.

Dr. John Cooley – Role: Tech Advisor. Leads FairwAI’s technical architecture, AI compliance engine, and research partnerships. Bio: Dr. John Cooley is a renowned technologist and serial entrepreneur who holds five degrees from MIT: B.S. in Electrical Engineering, B.S. in Physics, M.S. in Electrical Engineering, an Engineer’s Degree, and a Ph.D. in Electrical Engineering. He earned the David Adler Memorial Thesis Prize and the Morris Joseph Levin Award, recognizing his groundbreaking work in engineering and systems design. John was the CTO and later CEO of Nanoramic Laboratories, where he led the company to raise $44M+ in capital and developed cutting-edge battery and nanotechnology solutions. With 15+ years of experience at the frontier of hardware-software integration and AI systems, John brings a rare blend of research rigor and venture-scale execution. He now advises FairwAI on tech architecture, compliance logic, and academic research integration across MIT and international institutions.

Dr. Ryan Chin – Strategic Advisor, Urban Tech Mobility Systems Role: Ryan advises FairwAI on systems architecture, urban integration, and commercial strategy. He brings expertise in AI-driven mobility, sensor networks, and resilient infrastructure to inform how FairwAI scales across physical environments like hotels, hospitals, and cities.

Bio: Dr. Ryan Chin is a globally recognized expert in smart cities, urban mobility, and sustainable design systems. He co-founded Optimus Ride, a leading autonomous vehicle company spun out of MIT, where he served as CEO and led the development of self-driving tech for campuses, cities, and business districts. Prior to that, Ryan was Managing Director of the City Science Initiative at the MIT Media Lab, where he pioneered AI-integrated mobility systems and urban computing platforms. Ryan holds a Ph.D. and two Master's degrees from MIT in Media Arts and Sciences and Architecture. A serial entrepreneur and systems thinker, Ryan has advised Fortune 100 companies, U.S. government agencies, and international city governments on emerging technologies, sustainability, and mobility innovation. He is a frequent keynote speaker at Davos, Smart City Expo, and the UN. At FairwAI, Ryan supports us with building out our advisory team and the integration of computer vision, IoT, and AI fairness protocols into real-world infrastructure, ensuring our platform is adaptable, sensor-compatible, and aligned with the future of ethical urban technology.

This is a 12-week, full-time fellowship for someone who can build the technical backbone of a real applied AI platform.
You will own the AI surface of the system: the speech-to-text pipeline behind the documentation product, the clinical LLM that drafts SOAP notes, the coding model that proposes ICD-10-CM and CPT codes, and the eval harness that proves any of it actually works. The clinical data you train and evaluate on is sensitive, so the work is as much about rigor, reproducibility, and provenance as it is about model performance. You will work behind a strict inference gateway, not directly against vendor SDKs in product code.

What you'll do

Stand up and operate self-hosted inference for Whisper-large-v3 and the open-weights LLMs we depend on (Llama-3.1 family) on internal GPU infrastructure using vLLM.
Fine-tune the clinical note generator and the coding model on consented and licensed clinical corpora; manage data provenance, licensing, and de-identification meticulously.
Design the prompt templates and structured-output contracts that the inference gateway serves; treat the prompt registry as production code with reviews and versioning.
Build and maintain the evaluation harness: blinded comparison against human-authored notes, top-k agreement against coder ground truth, denial-risk model AUC, drift detection on live traffic.
Build and maintain the per-patient retrieval layer over the timeline using pgvector, paying close attention to retrieval failure modes that would lead to ungrounded clinical claims.
Partner with the clinical-services engineer on source-evidence enforcement so that no unsourced AI claim ever reaches a signed note.
Write and own the model cards, evaluation reports, and bias audits required for FDA SaMD pursuit on the relevant features.

What you bring

Hands-on experience fine-tuning and serving open-weights LLMs in production, not just notebooks. You have deployed something that real users hit.
Strong applied ML fundamentals — eval design, dataset curation, error analysis, statistical reasoning — beyond just model training.
Familiarity with structured-output techniques: function calling, constrained decoding, schema validation against model outputs, and the tradeoffs between them.
Experience with at least one ASR system (Whisper, Conformer, or comparable) and a working understanding of diarization and speaker assignment.
Working knowledge of vector retrieval (pgvector, FAISS, or similar) and a clear-eyed view of where naive RAG breaks down.
Healthy skepticism toward public benchmarks and comfort designing domain-specific evals when those benchmarks miss the point.
Clinical NLP experience or a serious appetite to develop it; familiarity with UMLS, SNOMED, ICD-10-CM, or CPT is a plus.

This role is a minimum of 30 hours a week in the summer or 15 hours a week during the school year. This role is unpaid, however you can do CPT/OPT or extenal school funding from school and foundations for our non profit.