Senior Machine Learning Engineer

This is a fully remote position, open to applicants in Brazil.

📋 Description

• You will serve as the key link between prototype development and production implementation.

• Your role will involve designing and constructing the machine learning engineering for the pricing engine, including inference serving, training pipelines, and feature engineering, ensuring that intricate models operate in real time with minimal latency on Ray.

• The emphasis will be on modeling and machine learning code; the MLOps/Platform team will manage the platform and runtime, and you will collaborate closely with them.

• For inference serving, you will design a pipeline for interconnected models using Ray Serve, focusing on model composition, low latency, and update strategies.

• You will develop training pipelines (Ray Train/Data), hyperparameter optimization (Ray Tune), and tenant-specific trained models with robust checkpointing for distributed training.

• In feature engineering, you will define and implement features in the feature store (Feast/Redis), ensuring consistency between training and production environments.

• Your responsibilities will also include implementing and optimizing components for linear programming and offline reinforcement learning within the pricing pipeline.

• You will monitor model quality by assessing drift from a modeling perspective, validating versions, and generating explainability reports (SHAP) in collaboration with MLOps.

• As a technical leader, you will act as a reference point, mentor others, and work with the team to determine feasible and scalable solutions.

• You will facilitate handoffs with data scientists, accept data from data engineers, and deliver it to the MLOps/Platform team for deployment and operational purposes.

⛳️ Requirements

• Demonstrated experience in deploying machine learning models in production environments.

• Proficiency in Python and solid software engineering principles (APIs, testing, clean code).

• Experience in inference serving and optimization techniques for achieving low latency.

• Familiarity with containerization tools (Docker) and MLOps processes (model registry, deployment).

• Comfort with AI-assisted development tools (e.g., Claude Code).

• Strong familiarity with the Ray ecosystem (Serve, Train, Tune, RLlib) is a significant advantage.

• Knowledge of feature stores (Feast) and scalable low-latency serving using Redis.

• Experience with optimization solutions (Gurobi, HiGHS) or real-time revenue management, along with offline reinforcement learning expertise.

• Experience in serving generative AI models (vLLM, LiteLLM) and working with multi-tenant architectures.

🏝️ Benefits

• Opportunity for remote work.

• Project duration of 6 months, with the chance for extension or potential internal hire.