Job Description
Position overview
The Head of AI/ML R&D decides, with evidence, whether an AI system works, how well, for whom, and at what cost. They come from data science and machine learning, and they build today with AI-native tools and agentic development.
The role owns the science of evaluation across every agent, the predictive and decision models behind the shared agents (churn, demand, payment recovery, fraud), and the experiment design that lets Finance count AI value once, against a holdout.
The role is hands-on. At least 50% of time goes to building models, evals and experiments; the rest to setting methods, reviewing, and growing a small team. This role makes what engineers ship measurable, cheaper and trustworthy, and supplies the models where an LLM alone is the wrong tool.
Key responsibilities
1. Evaluation science for every agent
o Set the enterprise eval methodology: golden-set design, sampling, labelling and inter-rater agreement, LLM-as-judge validation, slice analysis and statistical thresholds.
o Own the quality signals in the minimum telemetry contract: task success, groundedness, hallucination and escalation rates.
o Re-run evals per BU and tenant on every agent-core upgrade; catch regressions before customers do.
2. Predictive and decision models behind the shared agents
o Build and run the models the horizontal pods need: churn and retention uplift, payment-recovery propensity and timing, demand and capacity forecasting, later fraud and onboarding risk.
o Choose the right tool per task: classical ML, a small or fine-tuned model, retrieval, or a frontier LLM.
o Ship models to production on the golden path, with monitoring for drift and feedback loops.
3. Model economics
o Build the evidence behind model routing: accuracy, cost and latency per task, so routing is a measured decision.
o Drive cost per successful task down through distillation, caching strategy, smaller models and prompt and context optimization.
4. AI-native R&D and horizon scanning
o Prototype varied techniques (agent memory, planning, multi-agent patterns, synthetic data) against real Xplor problems.
o Turn what works into reference implementations in the golden path; retire what doesn't.
o Bring a monthly view of what has changed in models and methods, and what it means for our roadmap.
5. Data science for Data-as-a-Service
o Shape the statistical design of benchmarks and indices: cohort minimums, anonymization, and comparability across businesses.
o Lead entity-resolution modeling with the data team.
6. Raise the ML fluency of the whole R&D organization
o Teach evals, experimentation and probabilistic thinking through the AI-P/SDLC training and office hours.
o Pair with BU engineers on hard model problems in lighthouse BUs; review eval designs for R2 and R3 agents.
Hands-on expectations
• Writes production code, notebooks and eval harnesses; uses AI coding and agent tools daily at "platform contributor" fluency.
• Builds the first eval suite and the first predictive model for each golden-path exemplar (conversational analytics, booking agent, retention and churn).
• Reviews model and eval pull requests from pods and BUs.
• Decides on a demo, an eval or a metric, not a deck.
Qualifications
Required
• 12+ years in data science, machine learning or applied research, with 4+ years leading teams of scientists or ML engineers.
• Grew up in ML: has trained, evaluated, deployed and maintained models in production, not only consumed APIs.
• Has shipped LLM or agentic systems to production in the last two years, including the eval suite that gated them.
• Deep experimental design and causal inference: A/B tests, holdouts, uplift modelling, matched cohorts, and knowing when each is valid.
• Strong statistics: sampling, confidence intervals, calibration, multiple-comparison traps.
• Builds with AI-native tools today (agentic coding assistants, MCP, agent frameworks) and can show recent work.
• Writes production-quality Python and SQL; reviews others' code credibly.
• Has worked on SaaS or marketplace data at scale: multi-tenant, messy, many products.
• Explains model trade-offs and uncertainty to executives in business terms (revenue, margin, risk).
Preferred
• Churn, retention, propensity or pricing models in subscription or SMB software.
• Payments ML: recovery, dunning optimization, fraud or risk scoring.
• Fine-tuning, distillation or small-model work to cut inference cost.
• LLM-as-judge design and validation; red-teaming of model behavior.
• Snowflake (or similar) and feature stores; Datadog or OpenTelemetry for model monitoring.
• Familiar with NIST AI RMF, ISO/IEC 42001 and the EU AI Act, and with privacy rules for analytics across AU, NZ, US, UK and EU.
• Published research or open-source work is a plus, not a requirement.
Competencies and behaviors
• Evidence first. Brings a number with its error bar.
• Builder, not only researcher. Measures success by what reaches customers.
• Pragmatic. Picks the simplest model that clears the bar; says no to research that has no BU sponsor.
• Teacher. Grows ML judgement in engineers who don't report to them.
• Security- and privacy-minded. Treats training data, prompts and tool output as untrusted; no card data, credentials or customer data where it doesn't belong.
• Partner to BUs. Treats BUs as customers; works within the co-build-with-two-BUs rule.
Additional Information
All your information will be kept confidential according to EEO guidelines.
Before you apply
To be considered for employment, you must be legally authorized to work in the country where the role is advertised. This applies to remote roles too. Xplor does not sponsor visas, either at the time of hire or at any later time. We're also not able to offer contractor arrangements or Employer of Record engagements for this position at this stage.
Please apply through our careers portal or external job boards only. We're unable to upload emailed applications, as our process requires you to answer screening questions directly.
We also invite you to check out our Candidate FAQs for more information about our recruitment process xplor.com/recruitment-faqs/.
Interview Recording and Background Checks
As part of our hiring process, successful candidates will be asked to complete background and identity verification checks. Many of our roles are remote, and to protect both candidates and our business from recruitment fraud, we may record video interviews during the process. We keep any recordings and related data only for as long as the hiring process is active and do not retain them afterwards.
