Job offer · Research & AI · Fully remote, no location constraint · Permanent
AI research engineer (self-hosted LLMs, voice)
Advance the self-hosted language models that power the voice assistant and Nexum Health AI's agents: training, evaluation, production deployment. A role at the crossing point of applied research and AI infrastructure.
The role
Nexum's voice assistant and Nexum Health AI's agents rely on self-hosted language models, run in production in over 400 pharmacies. The role is to advance these models: evaluation, tuning, voice processing, and production deployment with the technical team.
This is not an isolated research role: you work with real data, concrete latency and reliability constraints, and a team that runs its own servers rather than depending on an external provider.
What you will do
- Evaluate, tune and advance the self-hosted language models used by the voice assistant and the AI agents.
- Work on voice processing: recognition, synthesis, latency, robustness to varied accents and environments.
- Deploy your results to production with the technical team, on infrastructure run in-house.
- Measure the real quality of agent responses and propose measurable improvements.
- Talk with the scientific board of pharmacists to ground technical choices in real usage.
What we are looking for
- Concrete experience in machine learning, ideally on language models or voice processing.
- Practice running and evaluating self-hosted models, not just third-party APIs.
- A taste for measuring before asserting: rigour comes before intuition.
- English for technical exchanges and scientific literature; French is a plus.
- The wish to see your work serve a real product, used every day by pharmacists.
- The models are trained and run on health data under HDS constraints: securing this data is part of the research work, not a side topic.
What we offer
- Fully remote, from any country; the team is already international.
- Self-hosted LLMs and servers run by a dedicated team, with no dependency on an external provider.
- An award-winning product, PharmagoraPlus 2026 Innovation Award, deployed in over 400 pharmacies.
- A research team being built and a scientific board of pharmacists that challenges the ideas.
The process, in four steps
- A conversation with the CTO.
- A technical interview on language models and voice processing.
- A live working session.
- A meeting with the team.
No quiz, no code link to provide: we prefer to see you work. Decision within 48 hours after the last step.
Apply
The form prepares your email, position already filled in: a few lines about you and what draws you here, a link to your work. The response comes from the person recruiting for the position.
A few lines about you and what draws you here; CV or online profile welcome. Address.
All positions
- Data analyst — acquisition and usage (data-driven marketing)
- Data engineer — health data platform (HDS)
- Data scientist — pharmacy stock and demand forecasting
- Senior backend engineer — pharmacy management software connectors
- DevOps / SRE engineer — HDS infrastructure and self-hosted LLMs
- Senior fullstack engineer — TypeScript
- React Native mobile engineer — Nexum Pharma and Nexum Pro
- Cybersecurity engineer — health data and compliance
- Product and regulatory pharmacist — scientific board and journeys
- PhD candidate in conversational AI — research team
- PhD candidate in AI applied to health data — safety and evaluation
- Product designer — counter and app interfaces
- Product manager — pharmacy and patient journeys
- Operations manager — pharmacy deployment and success
- Operations manager, Spain / Portugal — market opening
Come and write what comes next.
A conversation with the CTO, an architecture interview, a live working session — then a decision within 48 hours.
Fully remote · Team of eight · Everything written in-house