Senior ML Engineer for Real-Time Recommendations & LLM

Maropost; United States; Remote; USD 140,000 - 210,000; Full time

Job summary

Maropost is seeking a seasoned ML Engineer to own product-discovery ML components, including personalized search ranking, recommendations, and an LLM-assisted shopping assistant. You’ll train models, build in-production services, and keep systems running at scale for a global merchant base.

The role covers end-to-end ML pipelines, from embeddings to evaluation, with a small, fast team, real-time inference below 200 ms, and deployment ownership using containers and Kubernetes.

Qualifications

• 5+ years building ML systems with at least 3 years serving models in production.
• Strong Python skills with FastAPI or equivalent.
• Experience with vector databases (Qdrant) and ANN libraries (FAISS, Annoy, ScaNN, HNSWlib).
• Hands-on recommender systems or search ranking with embeddings.
• SQL proficiency; ClickHouse experience is a plus.
• Docker and Kubernetes; able to deploy and own your services.
• Experience running AB tests and interpreting results.
• Streaming systems experience (Pulsar, Kafka, Flink).

Responsibilities

• Design, train and ship recommendation and ranking models using embeddings.
• Build and operate Python services that serve them with low latency.
• Run batch pipelines to rebuild model artifacts across merchants.
• Design and evaluate AB tests; decide features based on evidence.
• Own deployments: containers, Kubernetes manifests, autoscaling, dashboards.

Skills

ML systems

Python

Recommender systems

FastAPI

Vector databases

Qdrant

FAISS

Annoy

ScaNN

HNSWlib

SQL

Docker

Kubernetes

AB testing

Streaming systems

Pulsar

Kafka

Flink

Tools

Docker

Kubernetes

ClickHouse

Job description

Maropost is seeking a seasoned ML Engineer to own product-discovery ML components, including personalized search ranking, recommendations, and an LLM-assisted shopping assistant. You’ll train models, build in-production services, and keep systems running at scale for a global merchant base.

The role covers end-to-end ML pipelines, from embeddings to evaluation, with a small, fast team, real-time inference below 200 ms, and deployment ownership using containers and Kubernetes.