AIML ENGINEER

What matters most

• Deliver production AI features

  Build and enhance AI Launchpad features through model integrations and prompt/workflow configurations. The posting requires experience building applied ML or LLM features in production software.

• Turn experiments into recommendations

  Run AI evaluations and experiments, then produce evaluation outputs, prototypes and technical recommendations for ADM.

• Improve grounding and reliability

  Build retrieval and data patterns that support feature accuracy, grounding and reliability, drawing on RAG, API integration and data-pipeline experience.

• Keep Launchpad interoperable

  Coordinate with the Marketplace POD’s Agentic AI Engineer on MCP-based tool and context patterns and emerging agent-to-agent patterns. MCP familiarity is preferred, not listed as required.

• Support feature stabilization

  Contribute technical documentation and analytics, and assist with testing, defect triage and stabilization alongside the Launchpad Tech Lead and Flex Specialist.

Likely challenges to think through

• Measuring improvements credibly

  Preparation hypothesis: a prompt or retrieval change may improve some outputs while weakening others. Prepare to explain how you would choose evaluation cases, compare results and turn findings into a technical recommendation; the posting does not specify an evaluation framework.

• Designing for evolving interfaces

  Preparation hypothesis: keeping Launchpad compatible as Marketplace MCP and agent-to-agent patterns mature may require balancing near-term delivery against flexible integration boundaries. Prepare an example of coordinating an interface change across teams with separate backlogs and sprint commitments.

• Balancing feature work and stabilization

  Preparation hypothesis: new model integrations and workflows may compete for time with testing and defect triage. Prepare to describe how you would isolate an AI-feature defect, assess its impact and decide whether to revise a prompt, retrieval pattern or integration.

Topics to prepare

• Production ML/LLM feature design

  Required experience includes building applied ML/LLM features in production software. Sketch a Launchpad feature from model integration through testing and stabilization, and explain where reliability risks arise.

• Retrieval and grounding

  The role builds retrieval and data patterns for accuracy, grounding and reliability. Compare two retrieval designs on how they handle irrelevant or missing source material, then explain the likely effect on generated answers.

• Model evaluation methodology

  Model evaluation is required, and the role produces evaluation outputs and technical recommendations. Define a small test set, success criteria and a method for comparing two prompt or model configurations.

• API integration and data pipelines

  A working understanding of API-based service integration and data pipelines is required. Diagram how source data would reach a retrieval-enabled feature, identifying integration failures and data-quality checks.

• MCP and agent interoperability

  MCP concepts, tool calling and agent orchestration are preferred; the role also coordinates on emerging MCP-based tool and context patterns. As preparation, trace a hypothetical tool request and identify the context and interface information Launchpad would need to exchange.

Skills to revise

• Python model integration

  Strong Python and experience with LLM SDKs or frameworks are required. Practice writing a small Python function using an OpenAI or Anthropic API, LangChain, or an equivalent to submit a request, handle an error and return a usable response.

• Prompt and workflow configuration

  Prompt engineering is required, and Launchpad work includes prompt/workflow configurations. Draft two prompt versions for one feature, specify their workflow steps and test how each handles ambiguous input.

• Retrieval-augmented generation

  Practical RAG experience is required. Build a small retrieval-and-answer prototype over sample documents, then inspect which retrieved passages support each answer and where grounding fails.

• Evaluation outputs and recommendations

  The role runs experiments and produces evaluation outputs, prototypes and technical recommendations. Run the same examples against two configurations, record results and write a brief recommendation tied to the evidence.

• Service and pipeline implementation

  API integration and data pipelines are required knowledge. Practice moving sample records through an API-backed ingestion step, validating fields and handling a failed request before the data reaches retrieval.

• Defect triage and technical documentation

  The posting calls for testing, defect triage, stabilization and technical documentation. Reproduce a faulty AI-feature output, record the input and configuration, isolate a likely failure point and write a concise issue note.

• Map evidence to the required experience

  Prepare a one-page map of dated examples showing your own contributions to production ML- or LLM-powered features, Python, LLM SDKs or frameworks, prompt engineering, RAG, evaluation, API integration and data pipelines. The posting specifies 4+ years of production feature experience; leave any unsupported area unclaimed.

• Rehearse a production feature walkthrough

  Select one relevant project and prepare a five-minute explanation of its model integration, prompt or workflow design, Python implementation, production constraints, your decisions and observed results.

• Sketch a retrieval design

  Practice a whiteboard design for an AI Launchpad-style feature that retrieves data and produces grounded answers. Show the data flow, retrieval choices, ways to check accuracy and reliability, and key failure cases; treat the design as an interview exercise, not a stated project requirement.

• Prepare an evaluation artifact

  Bring a concise example or proposed template with test cases, comparison criteria, evaluation outputs and a technical recommendation. Explain how you would investigate grounding failures; agent safety testing is listed as preferred exposure, not a requirement.

• Diagram an interoperability approach

  Sketch how a Launchpad feature could exchange tool and context information with MCP-based patterns and, as those patterns mature, agent-to-agent workflows. Identify assumptions and questions for the Marketplace team rather than presenting a particular integration as already implemented.

• Show how you stabilize and communicate a feature

  Prepare an example of a defect investigation or a proposed triage plan, including a reproducible test, useful analytics and a short technical note. Connect it to the posting’s testing, documentation and stabilization responsibilities.

• Write focused questions for the team

  Ask the Launchpad Tech Lead how evaluation results inform feature decisions, what retrieval reliability problems matter most, and how Launchpad and Marketplace coordinate shared AI patterns while maintaining separate backlogs and sprint commitments.