Machine Learning Engineer, CX Intelligence

Coinbase
Rio de Janeiro
Workplace: RemoteFull timeBRL 347,900 - 347,900 annuallyFunction: Data Science & Machine LearningExperience: 3+ yearsSkills: ["Communication","Stakeholder management","Responsible AI","Human oversight"]

Design and scale agentic AI systems for customer support, powering Coinbase Chat and Help Center. Build multi-agent workflows with LangGraph and Google ADK, integrate LLMs via Model Context Protocol (MCP), and create “LLM-as-a-judge” evaluation pipelines for production monitoring. Apply RAG, fine-tuning, and prompt engineering to keep responses accurate, grounded, and compliant, and ship resilient, low-latency Python microservices at Coinbase scale.

Loading

Loading job details...

Preparing the role view and application actions.

FursaFursa
Coinbase
Coinbase
1 month ago

Machine Learning Engineer, CX Intelligence

✓ Verified Job

Canonical indexed version, validated from employer's careers page.

Source: Company careers pageValidated by: Fursa AI
Last checked: 1 hour agoStatus: Live

Job Summary

Design and scale agentic AI systems for customer support, powering Coinbase Chat and Help Center. Build multi-agent workflows with LangGraph and Google ADK, integrate LLMs via Model Context Protocol (MCP), and create “LLM-as-a-judge” evaluation pipelines for production monitoring. Apply RAG, fine-tuning, and prompt engineering to keep responses accurate, grounded, and compliant, and ship resilient, low-latency Python microservices at Coinbase scale.
Location: Rio de Janeiro
Workplace: Remote
Employment Type: Full time
Job Function: Data Science & Machine Learning
Seniority: Mid level

Key Responsibilities

  • •Architect multi-agent systems using LangGraph and Google ADK to automate complex customer support workflows end-to-end.
  • •Build and scale MCP integrations connecting LLMs to internal Coinbase APIs, databases, and third-party tooling.
  • •Develop LLM-as-a-judge evaluation pipelines to monitor and improve non-deterministic AI agent performance in production.
  • •Implement RAG, fine-tuning, and prompt engineering to keep chatbot responses grounded, accurate, and compliant with Coinbase policies.
  • •Ship production-ready Python services that are resilient, low-latency, and able to handle Coinbase-scale traffic across asynchronous microservices.

Pay and Benefits

Salary: BRL 347,900 annually
Equity and Bonus:Equity
Perks:Health InsuranceDentalVision

Key Requirements

  • •3+ years building and scaling production ML/AI systems, including agentic “Loop” or “ReAct” approaches where AI takes autonomous actions.
  • •Deep understanding of the LLM lifecycle (context window, token optimization, structured output parsing with Pydantic/JSON mode) and multi-agent orchestration frameworks like LangGraph and Google ADK.
  • •Strong Python proficiency with experience in microservices architecture, asynchronous programming, high-throughput APIs, and vector databases (Pinecone or Weaviate).
  • •Ability to explain model behaviors and technical trade-offs to non-technical stakeholders across CX, Legal, and Product teams.
  • •Use generative AI responsibly with human oversight to deliver measurable improvements in workflow efficiency, cost, and quality.
Experience:3+ years
Skills:CommunicationStakeholder managementResponsible AIHuman oversight
Languages:English
Tech Stack:PythonLangGraphGoogle ADKModel Context Protocol (MCP)LLMsPydanticJSON modeRAGFine-tuningPrompt engineeringMicroservicesAsynchronous programmingHigh-throughput APIsPineconeWeaviateVector databasesLLM-as-a-judgeStructured output parsing

Company Brief

Coinbase
Operates a cryptocurrency exchange and financial services platform that enables buying, selling, storing, and earning crypto assets, institutional trading, custody, and developer tools for blockchain applications.
Industry: Trading Platforms
Company Size: Enterprise (1,001+ employees)
Revenue: USD 1B+
Growth: Public Company
Valuation: Public Company (Market Cap in USD)
Funding: IPO / Publicly Listed
Headquarters: San Francisco, United States
Founded: 2012
WebsiteLinkedIn