Level: Associate to Principal /Lead – title and scope matched to your experience

The Role

Our client is not a research lab publishing papers that sit on arXiv. They’re not a consultancy building POCs that never see daylight. They are putting machine learning into production at scale, where latency is measured in milliseconds and mistakes impact millions of users.

They’re hiring Applied ML Scientists across all levels to join a team that owns models end-to-end — from data to features to training to serving to monitoring. If you want your work to ship to real products instead of sitting in notebooks, this is the team.

What You’ll Work On

You will design, develop, train, and deploy ML models that power recommendation, personalization, computer vision, NLP, and forecasting systems. Your models won’t be experiments. They go live to web, mobile, and broadcast platforms within weeks, serving 10M+ users. You’ll solve messy, real-world problems like cold start, imbalanced data, 100ms latency budgets, regulatory constraints, and editorial policy. This isn’t Kaggle. You’ll work embedded with Product and Engineering, with direct access to the business problem and weekly reviews with the CTO. There is no “throw it over the wall to MLOps” — you instrument, deploy, and own what you build.

The Stack

You’ll work across PyTorch, TensorFlow, JAX, Transformers, and LLMs, plus RecSys, CV, and GNNs depending on the team. For MLOps, the client runs Kubernetes, Ray, Kubeflow, MLflow, Feast, Triton, and vLLM. Data runs on Spark, Kafka, BigQuery or Snowflake, and modern Vector DBs. All of this is on GCP or AWS at real-time scale. You don’t need every tool on Day 1, but you do need the curiosity and ability to go deep fast.

Who Thrives Here

Our client hires for slope, not just y-intercept. You’ll thrive if you’ve shipped ML to production and dealt with data drift, A/B tests, rollback plans, and the reality of users complaining when v2 is worse than v1. You think in trade-offs and can defend why XGBoost beat your Transformer to the board. You code for production — Python is fluent, you write tests, and you understand what “p99 under 200ms” means for model design. You’re product-minded and ask “should we build this” before “can we build this”. And you learn aggressively, because the field changes every 6 months and so does the roadmap.

A PhD or Masters in ML, CS, Stats, or Math helps, but equivalent experience wins too. Publications, Kaggle rankings, or a GitHub you built at 3am because it was fun all count. Domain experience in media, recsys, ads, or robotics is a plus, but not required.

Why Our Client

Your model will impact millions, not thousands. When it works, senior leadership knows your name. Code reviews are humbling, in the best way.

Think cold start for new content, multi-objective recommender systems, live AI moderation, and inference cost at broadcast scale. Not “predict churn on a 2017 dataset”.

Career optionality. Grow to Principal Scientist, move into Engineering Management, or pivot to product.

You won’t get trapped as a “researcher”. And finally, compensation that respects the craft. Competitive base, equity, and bonus, with GPUs on demand and budget for conferences. They don’t do “innovation days” because shipping is the innovation.

To Apply

Send your CV with two bullets on the most impactful ML system you’ve shipped, and one problem you’d love to solve for our client.

Our client is an equal opportunity employer and is committed to building an inclusive environment for all.

Only Shortlisted candidates will be notified