Senior Full-Stack Engineer (AI-First)

This role is suitable for a Senior Full-Stack Engineer with 5+ years of experience building React/TypeScript and Node.js applications, ideally with PostgreSQL or Supabase and cloud infrastructure such as GCP. They should be skilled in AI-powered development, including RAG, prompt chains, agentic workflows, AI coding tools, secure data handling, testing, CI/CD, monitoring, and performance optimization. The ideal candidate is familiar with owning features end-to-end, developing clinical or data-intensive products, designing AI pipelines and real-time backends, and working across architecture, deployment, documentation, and healthcare security requirements.

How would you design a React and TypeScript architecture for a clinical chat interface with streaming AI responses?

I would separate conversation state, streaming transport, and presentation components, using typed API contracts and resilient loading and error states. I would stream responses through an Edge Function or server-side endpoint, persist messages securely, and ensure the UI supports cancellation, retries, accessibility, and auditability.

How would you implement a retrieval-augmented generation pipeline for clinical literature search?

I would ingest and normalize documents, generate embeddings, store them in a vector-capable PostgreSQL setup, and retrieve relevant passages using metadata and similarity filters. The prompt would include grounded context, require citations, and enforce safeguards against unsupported medical claims.

How would you design a lab report OCR pipeline using AI?

I would validate and securely upload the document, run OCR, normalize extracted values and units, and pass the structured content through a model with a strict schema. The system should preserve the original source, expose confidence levels, require human review for ambiguous results, and log each processing step.

How would you implement medication interaction checking safely?

I would combine a trusted, versioned clinical interaction database with AI only for explanation and user-friendly summaries, rather than relying on the model as the source of truth. Results should include severity, evidence, affected medications, timestamps, and clear escalation guidance for clinicians or patients.

How would you secure patient data in a Supabase application?

I would use strict Row Level Security policies, least-privilege service roles, encrypted transport, and AES-256 encryption for sensitive data at rest where application-level protection is required. Secrets would remain server-side, access would be audited, and retention, deletion, backup, and key-rotation policies would support HIPAA-aligned controls.

How would you design Supabase real-time subscriptions for clinical data without leaking information?

I would authorize every subscription through database policies and ensure users can receive only records belonging to their permitted tenant, patient, or care relationship. I would minimize payloads, avoid broadcasting sensitive fields, handle reconnection and ordering, and test policies with unauthorized access scenarios.

How would you optimize a slow Supabase query or PostgreSQL-backed feature?

I would inspect query plans with EXPLAIN ANALYZE, review indexes and filters, reduce unnecessary columns and joins, and check whether Row Level Security policies are affecting performance. I would also consider pagination, materialized views, caching, and appropriate connection management before changing the schema.

How would you reduce cold starts and improve reliability in Supabase Edge Functions?

I would keep functions small, avoid expensive initialization and large dependencies, reuse clients where the runtime allows it, and move noncritical work to asynchronous jobs. I would add timeouts, retries with backoff, structured logging, metrics, and alerts for latency, failure rates, and provider degradation.

How would you test AI-powered features whose outputs are nondeterministic?

I would test deterministic components such as validation, permissions, parsing, and persistence with unit and integration tests, then use curated evaluation datasets for model quality. Assertions should focus on schema validity, grounding, safety, citation behavior, and regression thresholds rather than exact wording.

How would you use AI coding tools while maintaining code quality and security?

I would use them for exploration, scaffolding, refactoring, tests, and documentation while treating generated code as untrusted. Every change would pass review, type checking, linting, security scanning, and automated tests, with special scrutiny for authorization, prompt injection, data exposure, and clinical logic.

Premedice is an AI-first health technology company building secure tools for clinical communication, lab report analysis, medication interaction checking, and patient data management. As a Senior Full-Stack Engineer, you’ll own features end-to-end across React, TypeScript, Node.js, Supabase, PostgreSQL, Edge Functions, AI orchestration pipelines, and secure encryption infrastructure. This is a fully remote, global, full-time role offering $80K–$100K plus equity, with the opportunity to use cutting-edge AI coding tools and shape products at the intersection of healthcare and artificial intelligence. Apply to make a meaningful impact in a small, high-ownership team while building the future of secure, AI-powered medicine.