Key responsibilities
• Translate business processes into AI use cases, establish functional and quality baselines, and
define testable acceptance criteria with customer stakeholders.
• Develop data ingestion, parsing, chunking, embedding, indexing, retrieval, and reranking pipelines.
Preserve document permissions and manage data updates and traceability.
• Build LLM applications and agents that use enterprise APIs and tools. Implement state handling,
retries, timeouts, structured outputs, and recovery from partial failures.
approved components where they meet deployment and application requirements.
• Create evaluation datasets and regression tests for factual grounding, task completion, retrieval
quality, and tool correctness. Investigate poor responses using traces and reproducible examples.
• Implement safeguards for prompt injection, unauthorized data access, unsafe tool execution, and
sensitive information exposure. Define approval steps for consequential automated actions.
• Containerize and deploy applications through CI/CD, configure monitoring, and optimize latency and
operating cost while preserving quality and maintainability.
• Work with platform, SRE, and data teams to resolve production issues. Document architectures,
integration contracts, test results, and operating procedures for customer handover.
Required experience and skills
• 7+ years in software engineering, applied machine learning, enterprise integration, or related delivery
roles, including hands-on generative AI application development.
• Strong Python, REST API, testing, source control, and debugging skills, with experience delivering
maintainable production software.
• Practical knowledge of LLM behavior, prompt design, embeddings, vector databases, retrieval
pipelines, and evaluation methods.
• Experience building agent workflows with tool calling, explicit state, access boundaries, and error
recovery using an appropriate orchestration framework.
• Working experience with PyTorch or TensorFlow and the ability to integrate model services into
containerized applications.
• Ability to diagnose failures across data pipelines, retrieval, model endpoints, application logic, and
infrastructure dependencies, and communicate findings clearly to customers.
