Clinical language & reasoning
Medical question answering, clinical NLP, and reasoning across longitudinal patient records.
Healthcare AI · Models, evidence, and connected care
Chandra Vikram Singh
I work on AI that can make sense of clinical information and fit into the systems people use to deliver care.
Healthcare AI engineer · MS in Health Informatics, Indiana University · Background in pharmaceutical sciences
About me
My route into AI began with pharmaceutical sciences and continued through graduate study in health informatics. That combination shapes the questions I bring to engineering: What information is missing? What evidence supports the answer? How will the result reach a patient or clinician?
I work across clinical language models, retrieval, interoperability, and applications for patients and care teams. I’m especially interested in the connections between them: grounding a model in the right evidence, carrying context from the health record, and making its uncertainty visible to the people using it.
Healthcare is my main focus, and I also explore broader applied AI. I value systems whose behavior can be inspected, tested, and improved.
What I’m exploring
Retrieval & knowledge graphs
Clinical RAG, guideline retrieval, patient context, and whether an answer’s citations actually support its claims.
Fine-tuning & alignment
Adapting open language models for clinical tasks and evaluating what changes after training.
Interoperability
HL7 FHIR, SMART on FHIR, and the infrastructure connecting AI applications with health records.
Agents, voice & telehealth
Clinical workflows, conversational interfaces, and useful handoffs between AI systems and care teams.
Evaluation & safety
Grounded benchmarks, privacy, uncertainty, failure analysis, and keeping clinicians involved.
Selected work & notes
Qwen3 fine-tuning experiment ↗
A public Qwen3 model adapted from Unsloth’s 4-bit Qwen3-14B checkpoint, trained with Unsloth and Hugging Face TRL.
Carepal ↗
A patient-facing healthcare application exploring AI assistance, patient-reported outcomes, and connections to clinical data systems.
Clinical AI working notebook ↗
Project ideas and preliminary manuscripts spanning FHIR infrastructure, clinical benchmarks, voice agents, knowledge graphs, retrieval, and alignment.
The notebook describes work in progress. Project proposals, evaluation plans, and targets should be read in that context.