Full Stack Engineer

Full Stack Engineer (with Infrastructure)
About Chattrly
Chattrly is a healthtech startup building the health intelligence ecosystem for patients and
clinicians. Two interconnected products, one thesis: Patient data in India is scattered across
thousands of labs, hospitals, and apps that don't talk to each other. We're closing that gap on both
sides — giving patients a unified view of their own health, and giving clinicians the intelligence layer
that turns fragmented data into instant, contextual insight. Every response we generate has to be
grounded, verifiable, and clinically safe — there is no room for hallucination when a doctor is making
a decision.
What You'll Actually Work On
You'll move across the full stack and occasionally into infrastructure. Concretely:
Backend
 Build and extend API endpoints that power the clinical Q&A experience and the patient app
 Work inside our AI orchestration pipeline — the layers that turn a clinician's free-text
question into a grounded, structured, trustworthy answer
 Design and evolve database schemas, write migrations against live tables, and ship them
safely to production
 Use caching and async task queues for background work like document processing, OCR, and
summarisation
AI / LLM Pipeline
 Integrate, tune, and evaluate calls to large language models — routing, structured output,
narrative generation, and verification
 Work on our hallucination defence layer: deterministic grounding checks plus model-based
gates that catch hallucinations before they reach the clinician
 Help with prompt engineering — we treat prompts as code: versioned, evaluated, and
deployed deliberately, not edited live
 Build evaluation harnesses so we can measure regression every time we change a prompt or
swap a model
Mobile & Web
 Ship user-facing features across our patient mobile app (iOS + Android) and our clinical web
platform — onboarding, document uploads, the chat surface, and more
 Work with native device capabilities like the camera, document scanning, and secure on-
device storage
 Debug platform-specific issues across iOS, Android, and the web — cold-boot lifecycle,
modal behaviour, keystore quirks, and the long tail of cross-platform weirdness
Infrastructure (occasionally)
 Help maintain our cloud deployments and the services that back them
 Write and review container builds, debug runtime issues, tune concurrency and timeouts
under real load
 Touch our deploy scripts (idempotent, atomic-swap, auto-rollback on smoke-test failure)
across staging and production
 Work with secrets management, server-level configuration, and database operations against
live environments
Observability & Quality
 Instrument features with structured logging — every model call, every critical path, with
cost, latency, and error telemetry feeding our dashboards
 Write unit, integration, and end-to-end tests; our pipeline has an e2e harness that exercises
real clinical scenarios
 Participate in code reviews. Every PR goes through a structured review covering correctness,
clinical safety, security, scalability, and data completeness before merge
You're a Good Fit If You
 B.Tech in CS and have 2-3 yrs of experience
 Are comfortable writing both a backend language and a frontend language — not just one.
You'll genuinely move between them in the same week
 Have deployed something — even a side project — to a cloud provider. Anything real, end-
to-end
 Can read a container build file and an orchestration config without panicking
 Understand REST APIs, relational databases, caching, and the basics of async / background
jobs
 Are curious about how LLMs actually work in production — context windows, structured
output, grounding, the latency-vs-quality tradeoffs. You don't need to be an ML engineer,
but you should want to understand the pipeline
 Can drop into a large unfamiliar codebase, read tests, search aggressively, and find your way
around without being hand-held
 Care about correctness. In our domain, "it kinda work" is a bug
 Remote-first, async-first. We respect your hours and how you work. Output over seat time

Work arrangement
No