Design, train, and deploy ML/LLM models that automate catalogue enrichment, moderation, and categorisation at scale, owning accuracy, latency, and cost in production.
Drive the transition from frontier API models to fine-tuned open-weight models: build data curation and fine-tuning pipelines, run quality comparisons, and take winners into production.
Design and build agentic AI systems for catalogue automation, covering multi-step workflows, tool use, guardrails, and the evaluation harnesses to prove they work.
Build evaluation and experimentation infrastructure: offline benchmarks, regression suites, LLM-as-judge pipelines, and A/B tests tied to business metrics.
Own the serving and cost story for self-hosted models, including quantisation, throughput tuning, GPU utilisation, and build-versus-buy decisions.
Collaborate cross-functionally with Software Engineers, Data Scientists, Product Managers, and the ML Platform team to productionise and monitor ML solutions.
About You
Proven experience building and shipping ML systems in production at scale, ideally in consumer-facing product environments at a technology company.
Demonstrated track record taking LLM-based systems to production, including prompt/model iteration, evaluation, guardrails, observability, and cost management.
Hands-on experience fine-tuning and serving open-weight models (e.g. LoRA/QLoRA, SFT, preference optimisation), including building training data and managing production serving.
Deep expertise in NLP or recommendation systems, with models that have moved a business metric.
Strong engineering fundamentals: mastery of Python and SQL, clean code practices, production experience with a deep learning framework (PyTorch, TensorFlow, JAX, or Triton), and experience with modern ML tooling and cloud infrastructure (AWS, SageMaker, Airflow, Docker).
Strong product sense and proactive ownership, with the ability to turn ambiguous problems into measurable ML solutions and drive them from discovery to production impact.
Linked-in
Google