About this role.
Sword is seeking a Europe-based Senior ML Engineer to build and operate production AI systems for a clinical-care platform. The role centers on reliable agentic LLM workflows, including retrieval, tool use, orchestration, evaluation harnesses, monitoring, and continuous improvement. The engineer will own projects from ambiguous exploration through deployment, while selecting techniques such as prompting, RAG, distillation, or fine-tuning based on evidence. Success requires strong production engineering judgment and close collaboration with product, clinical, and engineering stakeholders in a high-stakes healthcare environment. The position also includes code review, knowledge sharing, and mentoring of less-experienced engineers.
Role DNA
A quick view of the complexity, pace, ownership and collaboration implied by the job description.
Job Complexity
5/5Pace & Pressure
5/5Autonomy Level
5/5Communication Load
5/5Salary analysis
Estimated compensation compared with the broader US market for similar roles.
Core skills
Skills and capabilities most closely associated with this opportunity.
Sample interview questions
I would explain the user problem, system architecture, model and retrieval choices, launch criteria, and production metrics such as task success, groundedness, latency, cost, escalation rate, and user outcomes. I would also describe the monitoring and iteration process used after launch.
I would begin with a representative, versioned evaluation set covering routine, edge-case, adversarial, and safety-critical scenarios. The framework would combine deterministic checks, task-specific quality metrics, structured LLM-as-judge scoring, expert clinical review, and regression gates that prevent releases when quality or safety declines.
I would start with clear error analysis and baseline measurements. Prompting is appropriate for fast iteration and well-supported general capabilities; RAG is useful when answers require current or proprietary evidence; fine-tuning can improve repeatable behavior or specialized tasks with sufficient quality data; and distillation is useful when lower cost or latency is needed after establishing a stronger teacher system.
I would constrain tool permissions, validate inputs and outputs, use structured schemas, require confirmation for consequential actions, and implement timeouts, retries, idempotency, and audit logging. I would also use policy checks, human escalation paths, simulation testing, and production monitoring to detect unsafe behavior early.
I would frame the decision in terms of user and clinical impact, expected benefits, known failure modes, uncertainty, and operational cost rather than model terminology alone. I would present measurable options, recommend a safe rollout plan, and make explicit which risks require stakeholder or clinical-owner approval.
Since 2020, Sword has expanded across Musculoskeletal, Women’s Health, Cardiometabolic, and Mental Health, and is now moving beyond the session to a fully AI-native, 24/7 care program that brings physical activity, therapeutic exercise, psychotherapy, nutrition, and behavior change into one connected experience. More than 1 million members across three continents have completed over 15 million AI sessions, helping 2,000+ enterprise clients avoid more than $1 billion in unnecessary healthcare costs. Backed by 59 clinical studies, 43 patents, and more than $500 million raised from leading investors including Khosla Ventures, General Catalyst, and Founders Fund, Sword is defining a new standard for healthcare.
AI Proficiency at Sword
Explorer (Level 1) — Uses AI daily to boost personal productivity
Builder (Level 2) — Creates workflows and tools that elevate the whole team
Integrator (Level 3) — Embeds AI into products and processes at scale
AI fluency is a core expectation at Sword. Every candidate is assessed against our three-level framework — be ready to share real examples of how AI is already part of how you work.
Every hire must demonstrate at least Level 1. The expected level will vary depending on the seniority of the role.
What you’ll be doing:
- Own ML projects end-to-end: take problems from exploration through production deployment and keep iterating once real users are on them;
- Build agentic LLM systems: design multi-step workflows with tool use, retrieval, and orchestration, and make them reliable enough for clinical settings;
- Treat evaluation as core engineering work: build eval sets, offline and online harnesses, LLM-as-judge pipelines with human review, and regression tests that catch quality drops before they reach users;
- Improve model quality with whatever fits the problem: prompting, retrieval, distillation, or fine-tuning, chosen on evidence rather than habit;
- Work across the full AI stack: data prep, model adaptation, serving, monitoring, and the feedback loops that keep systems improving in production;
- Partner with Product, Clinical, and Engineering: translate clinical requirements into technical decisions and surface tradeoffs early;
- Help the team get better: review code, share what you learn, and mentor engineers earlier in their careers.
What you need to have:
- Experience shipping ML systems to production that people actually depend on;
- Hands-on LLM work in production: prompting, retrieval, tool calling, and agent-style workflows;
- A rigorous approach to evaluation: you’ve built eval datasets and frameworks, and you can tell a real improvement from noise;
- Strong ML fundamentals: you know which approach fits which problem and can reason clearly about tradeoffs;
- Comfort with ambiguity: you’ve taken loosely defined problems and turned them into something running in production;
- Solid engineering skills: production-quality code, familiarity with distributed systems, and the patience to debug messy ML pipelines;
- Clear communication with both technical and clinical stakeholders.
- Experience with fine-tuning or preference optimization (RLHF, DPO, or similar);
- Healthcare AI, or other high-stakes domains where errors carry real cost;
- Built agent frameworks or evaluation tooling from scratch;
- Open source contributions, technical writing, or other knowledge sharing.
Bonus points
These compensation bands are just the starting point. Once someone joins and proves they’re outlier talent, we adjust quickly to ensure their compensation aligns with their impact.
Our job titles may span more than one career level. Actual pay is determined by skills, qualifications, experience, location, market demand, and other factors. Compensation details listed in this posting reflect the base salary and any potential variable, bonus or sales incentives, and the Company’s estimation of the value of private company stock options, if applicable. The pay range is subject to change, future value of company stock options is not guaranteed, and compensation may be modified in the future. In addition to our total compensation, Sword offers a number of benefits as listed below.
Annual salary information is not provided for this position. Explore salary ranges for similar roles in our Salary Directory ›
Upload your resume now
- To unlock remote work opportunities and be discovered by global employers.
This job listing has been manually reviewed by the Jobicy Trust & Safety Team for compliance with our posting guidelines, including verification of the company's legitimacy, accuracy of job details, clarity of remote work policy, and absence of misleading or fraudulent content.
