CTO / Co-Founder

About the team

Clinical examination is one of the few remaining skills that cannot be learned from a textbook. It is learned by talking to patients, and access to that practice is rationed by faculty time, physical space, and geography. Klinico builds AI-powered clinical simulation so that a medical student's hundredth practice conversation costs the same as their first.

Our simulated patients are in use today by medical students in Palestine and Jordan. We are a founding team of one, and this is the second seat.

About the role

You will own the two systems that constitute the product: the simulated patient a student talks to, and the assessment engine that scores what they said.

Neither exists in a durable form yet. There is no backend service, no model orchestration, no retrieval layer, no evaluation harness. You will make those architectural decisions, ship them, operate them, and be accountable for their behaviour in front of the clinicians and students who use them.

This is a co-founder role, not a first-engineer role. You will set technical strategy, make hiring decisions, and decide what we do not build.

Responsibilities

• Design and ship the simulated patient system: a model that holds a hidden clinical history, discloses it only in response to appropriate questioning, maintains character under adversarial input, and calibrates difficulty to the student's training level.
• Build the assessment engine that converts a free-form conversation transcript into per-item scores and actionable feedback against a standardised clinical checklist, with reasoning defensible enough to survive challenge from a student and a faculty examiner.
• Own the retrieval architecture over our case library and supporting clinical material: vector store selection, embedding model, chunking strategy, and the latency/quality tradeoff on reranking. Retrieval quality here is directly measurable as a patient that does not fabricate symptoms and a grader that can cite the criterion it scored against.
• Build and maintain an evaluation harness so that prompt, model, and retrieval changes ship on evidence rather than on demo performance.
• Build the backend from zero: FastAPI service, case and session data model, streaming, queues, and background processing.
• Own cost and latency as product requirements. A single session is many model calls; both the unit economics and the time to first token are user-visible.
• Own safety and scope. A simulated patient must never provide real clinical advice, and the system must never process real patient data.
• Own the operational surface: deployment, CI, observability, backups, and secrets management — with the judgement to keep those choices boring.

What you would work on

• A completed front end: React 19, Vite, Tailwind v4, React Router, with authentication and onboarding shipped.
• Supabase, with Postgres row-level security as the authorisation boundary.
• No backend service, no orchestration, no retrieval, no grading pipeline, no evals. These are open decisions and they would be yours.

Minimum qualifications

• Experience shipping an LLM-based product to production and operating it over a sustained period. Prototype and demo work does not meet this bar.
• Strong Python, and production experience with FastAPI: async endpoints, streaming responses, Pydantic models, background tasks, and dependency injection.
• Experience designing, deploying, and debugging RAG systems in production, including chunking strategy, hybrid search, reranking, and evaluating retrieval independently of generation.
• Experience operating at least one vector database in production (pgvector, Qdrant, Weaviate, Pinecone, or equivalent), with a substantive account of the selection tradeoff and where it hurt.
• Experience with the broader applied LLM toolkit: structured outputs, tool and function calling, context design, prompt caching, model routing, agentic loops, LLM-as-judge and its failure modes, and the conditions under which fine-tuning outperforms prompting.
• Experience with PostgreSQL and schema design, treating authorisation as a design-time concern.
• Experience carrying features through a full lifecycle: user research, scoping, launch, measurement, and the decision to iterate or deprecate.
• Working proficiency in React where the work requires it.

Preferred qualifications

• A clinical background, or direct experience of the examination format we simulate.
• Professional proficiency in Arabic and English.
• Experience with realtime or voice interfaces.
• Experience building review and annotation tooling for domain experts.
• Experience as a founder or early engineer at a pre-product-market-fit company.

What we offer

• Equity. Meaningful co-founder equity on standard vesting with a cliff, discussed in the first conversation and documented before you start.
• Users. A product with students in it now, and direct access to the medical students and faculty who use it.
• Autonomy. Full ownership of technical decisions, and a co-founder who will not pretend to make them for you.
• Location. Remote within approximately three hours of Palestine time, or in Ramallah if you prefer an office.

On compensation: this role is equity-only. Klinico is pre-revenue and pre-funding, and there is no cash salary until that changes. We are stating this in the posting rather than in the third conversation. If you need a salary now, this is not the right time to join us.

Interview process

1. A 30-minute conversation about the product and what you would change about it.
2. A scoped, paid piece of real work, or an in-depth technical walkthrough of a system you have built and operated.

How to apply

Apply through this post, or read the extended version and send your CV at https://klinico.co/careers/cto

Include a link to something you have built and operated in production. We weight that considerably more heavily than a polished CV.

Klinico is an equal opportunity employer. We evaluate all applicants without regard to race, religion, national origin, sex, gender identity, sexual orientation, age, disability, or veteran status.