AI/ML Engineer — Agentic Systems (ID 1423)

Design and develop agentic AI systems that can verify their own outputs and recover from failures. Develop and optimize LLM-based agent workflows, including prompt design, tool use, structured outputs, retrieval, and context management. Write and review SQL queries against undocumented production database schemas and determine whether the available data supports specific conclusions.

Understand client business contexts and identify when AI-generated results may be incorrect or unreliable. Build evaluation frameworks by defining success criteria, implementing measurements, and monitoring system accuracy. Investigate cases where systems complete successfully but generate valid-looking yet incorrect outputs.

Evaluate the effectiveness of deterministic checks, safeguards, retries, and validation mechanisms. Work directly with business users to convert vague issues such as “this looks wrong” into specific, reproducible technical problems. Diagnose root causes using actual error data and implement appropriate fixes. Deploy and maintain reliable production agentic AI systems and improve their post-launch reliability.

Work independently on technically ambiguous problems and continuously learn new AI tools and techniques.

Requirements

Hands-on experience in LLM engineering and Agentic AI, beyond basic chatbot development. Strong understanding of structured outputs, tool calling, retrieval, context management, and prompt engineering. Ability to design prompts as versioned engineering artefacts rather than ad-hoc prompts. Strong SQL and database skills, particularly the ability to work with inconsistent or undocumented production schemas.

Strong analytical and debugging skills, with the ability to identify the actual root cause from error data. Ability to evaluate AI system outputs and verify claims before accepting or presenting them. Experience building AI/LLM evaluation frameworks, success criteria, measurement systems, and reliability metrics. Ability to understand business requirements and translate user-reported problems into reproducible technical issues.

Strong technical judgment regarding safeguards, validation checks, retries, and cost-versus-accuracy trade-offs. Experience with multi-tenant system architecture and security would be an added advantage. Strong independent problem-solving ability and willingness to learn rapidly in a fast-changing AI environment. A formal degree or fixed number of years of experience is not mandatory; the JD emphasizes demonstrated technical judgment and practical evidence of capability.