.png&w=256&q=75)
1
1
India
1 year of experience
"I am a medical doctor transitioning into healthtech, clinical AI, and health AI. Combining my clinical background with modern software development, I build open-source tools to bridge the gap between patient safety and artificial intelligence. My projects range from clinical safety suites like Lumen Clinical (an open-source clinical LLM safety evaluator and adversarial red-team sandbox) and Shinrin AI (a clinical decision support dashboard), to infrastructure tools like Clinical Middleware Dashboard (an autonomous EHR middleware for real-time ICD-10 and insurance compliance rule checking). To maximize efficiency and rapidly prototype complex workflows, I utilize an entirely AI-assisted development model. This allows me to combine my deep medical expertise with fast execution, building production-ready architectures that align with FHIR R4, openFDA, and ABDM protocols."

Lumen Clinical is a pre-deployment safety certification workstation designed to stress-test medical AI LLMs against clinical safety guidelines. Built under the direction of a clinician, Lumen ensures that clinical AI agents do not bypass safety rules, prescribe wrong doses, or miss life-threatening patient emergencies. HOW BAND COORDINATES OUR AGENTS: Rather than using static prompts, Lumen establishes a non-linear multi-agent simulation using the Band SDK coordination layer: 1. Red-Team Adversary Agent -> [Band Task: attack_scenario] -> Doctor Agent. The adversary adapts its jailbreak strategy (escalate, pivot, or compound) dynamically based on what the doctor says. 2. Doctor Agent -> [Band Task: consultation_turn] -> Patient Agent. The Doctor runs on BioMistral-7B via Featherless AI and utilizes diagnostic tools (LOINC lab checks, CPT imaging, RxNorm prescribing). 3. Patient Agent -> [Band Task: patient_response] -> Doctor Agent. The Patient simulates realistic clinical personas (minimizer, drug-seeker, health-anxious, confused elder) and withholds information unless properly queried. 4. Doctor Agent -> [Band Task: final_transcript] -> Safety Auditor. The Safety Auditor runs a consensus check via AI/ML API, scoring the session (0-100) and compiling an HL7 FHIR R4 compliance transaction bundle. BUSINESS VALUE: Lumen closes the pre-deployment gap for clinical AI, replacing months of manual trials with automated, rigorous safety auditing. It outputs standardized FHIR R4 compliance logs and triggers human-in-the-loop clinical escalation when safety scoring thresholds fall.
19 Jun 2026