We're hiring a Head of Engineering to own the technology end to end. You will be the senior-most technical leader for Elyx, responsible for architecture, delivery, reliability, security and healthcare-grade compliance, and the engineering team. This role builds the engineering capability that lets Elyx find what works efficiently: high-throughput prototyping, instrumented so the team learns from what clients do, not only from what they say.
You'll set technical direction, hire and grow the team, and stay hands-on in the hardest technical and architectural decisions. This is a high-ownership role for someone who has built complex, AI- and data-driven systems before and wants to do it in a domain that matters.
The work is held to two standards at once: it has to be sound, and it has to change what a client actually does. Sound means evidence-based and reasoned from mechanism—and, because no two clients are the same, tested against the individual rather than assumed from the population.
Key Responsibilities
Own the architecture, delivery, reliability, and security of Elyx. You will collaborate with the product team to understand user needs, set the engineering roadmap, and lead the team of engineers to execute the roadmap in a timely manner. You will also be responsible for building an engineering culture leveraging the changing AI landscape. The core components of the Elyx product are:
- The reasoning engine. Elyx runs continuously across everything known about a client—labs, biomarkers, wearable streams, imaging, meal photos, self-reports. The engine identifies the real patterns in one client's noisy data while excluding the false ones, and monitors and validates the evolving scientific literature for the findings relevant to that client.
- The econometric search engine. This is the module to use the client’s medical and health telemetry to search against the pool of scientific knowledge, biased by the client’s preferences, to generate a list of interventions and diagnostics that should be run to maximize the client’s health. The initial focus will be on improving existing health, but over time, this system will also be useful in helping analyze options for sickcare for our clients.
- The data model and capture underneath it. The model that turns those varied inputs into one coherent, queryable picture of the client, with a confidence level on each item. Includes capturing the reasoning specialists use when they design a client's health strategy, in structured form, so the engine can later learn from it.
- The client app. The surface the client and their concierge work in, and the team's primary means of learning what works: prototypes wired to record what clients actually do, so the team learns from behaviour rather than only from what clients report. The app's subject is communication—how it asks questions, how it shows progress, the tone it uses.
- The provider app. The tool the delivery providers and the concierge work in—personal trainers, physiotherapists, nutritionists, coaches, including freelance and overseas providers who pick up a client while they travel. It gives a provider the prepared brief on what to work on for that client and their current state, so a provider new to the client can deliver against the plan without assembling context, and captures the data from each session back into the system.
- The decision pipeline. When the system proposes a change to a client's plan, the change carries a prediction of what it should achieve. How rigorously that prediction is measured is negotiated with the client. Once shipped, the change is tracked to its outcome and the result feeds back. This discipline is designed with the specialist providers team.
- The data and access layer. Confidential health data where access is not a flat on/off list——a concierge, a specialist, and an external partner each see a different, scoped slice, bounded by client consent. Least-privilege access shaped by role and situation.
- The healthspan coach. Even though the client might know what to do to maximize their health, working through the list of diet, sleep, movement, etc. goals is not easy. We plan to anthropomorphize the AI agent into a healthspan coach to “brain hack” the client into adherence. This is a long term requirement that is part of our vision of where healthspan management will go.
Required skills and experience
- Engineering leadership. Proven track record leading engineering teams (10 to 25) as the senior-most technical owner of a complex production platform, responsible for architecture, delivery, reliability.
- Building the team. Proven track record of recruiting top-tier engineering talent for early-stage startups and building high-performance, motivated teams
- Hands-on. Actively contributes to the codebase (using AI or otherwise), including code reviews, and builds AI tooling to enhance team productivity; Owns the hardest technical problems directly. This is a building and managing role, not purely managing.
- Deep knowledge engineering. Demonstrated experience turning unstructured, high-stakes content into structured, queryable, evidence-graded knowledge that professionals act on. Required across two areas: a body of scientific literature, and one subject's noisy longitudinal data. Plus demonstrated experience structuring messy input at the point of capture.
- Applied AI / ML systems. Demonstrated experience building and shipping LLM/ML systems in production, with real evaluation, observability, and accuracy and hallucination control.
- Instrumentation-first. Demonstrated experience building products instrumented to learn from user behavior.
- Sensitive data access. Built systems where access to sensitive data is governed by role and situation—scoped slices, not a flat permissions scheme.
- Cloud and infrastructure. Strong modern cloud engineering (AWS or equivalent), CI/CD, and reliability practices.
- [Strong plus] Clinical data. Hands-on experience ingesting and structuring clinical data—labs, biomarker panels, imaging, EHR and external specialist records (e.g. FHIR / HL7)—into a form a system can reason over, with PHI handling under a real data-governance regime.
