Job Description & Details
This is a hands-on Machine Learning role focused heavily on building out the production ranking, retrieval, and personalization engines that drive modern AI applications. If you enjoy sweating the details of search relevance, managing embedding spaces, and pushing models all the way to production rather than just keeping them in a Jupyter notebook, this gig is worth a serious look.
What You'll Actually Be Doing
Expect to spend your days knee-deep in model optimization, tweaking classification and ranking pipelines, and making sure your inference latency doesn't spike under real user load. You'll be bridging the gap between raw user behavior data and efficient model execution—meaning you'll build evaluation frameworks, handle hybrid search strategies, and continuously debug why a particular reranking model is dropping precision in production. It's an end-to-end ownership gig where you build it, ship it, and own its reliability.
The Core Tech Stack
Python is your daily driver here, paired heavily with PyTorch or JAX for building out modern neural architectures. The team heavily relies on embeddings, vector search, and custom ranking models to drive personalization. You'll need a solid grasp of how to build and maintain end-to-end ML inference pipelines, balancing strict production constraints like inference cost and latency against model accuracy.
Interview Expectations
Be prepared for a brutal 60-minute coding assessment right out of the gate, so keep your data structures and algorithmic chops sharp. Beyond that, expect technical interviewers to press you on production failure modes—specifically, how you would diagnose a sudden drift in search relevance or optimize a slow-running retrieval pipeline without blowing up your infrastructure costs. They want to see practical engineering pragmatism over textbook theory.
Application Advice
When updating your resume for this role, drop the generic buzzwords and focus strictly on your production deployments. Explicitly mention terms like "vector search", "ranking models", "inference pipelines", and specific frameworks like PyTorch or JAX so your application sails through the ATS. Highlight metrics-driven impact from past roles—tell them about the latency you shaved off, the precision gains you engineered, or the scale of the production datasets you managed.