Senior Evaluation AI Engineer

We are seeking a Senior/Staff Evaluation AI Engineer to anchor the rigorous benchmarking, validation, and safety of our next-generation Agentic AI platform. Our platform automates highly complex enterprise workflows for corporate real estate (CRE), including automated lease abstraction, predictive portfolio optimization, and intelligent space management. This role sits at the intersection of Quality Engineering, Software Engineering, and AI/ML. The ideal candidate will have strong hands-on automation skills and experience testing LLM-, RAG-, and agent-based applications, with the ability to translate AI quality expectations into measurable evaluation criteria and automated regression suites. The AI Evaluation Engineer will partner closely with Engineering, Product, Data, and Quality Engineering teams to ensure AI capabilities are measurable, explainable, reliable, and production-ready. Duration: 6 months but high chance for prolongation in case of good performance, as this is long term role Key Responsibilities Architect Evaluation Frameworks: Design, implement, and scale custom automated evaluation pipelines for multi-agent LLM systems, focusing on intent-classification, tool-calling correctness, state transition integrity, checkpointer manipulation and multi-turn memory replay. Define Domain-Specific Metrics: Develop advanced custom metrics (incorporating G-Eval, LLM-as-a-judge architectures, and deterministic verification methods) to evaluate domain-specific tasks like complex document parsing (leases, invoices) and real estate financial modeling. Establish Guardrails & Safety Protocols: Implement defensive alignment testing, adversarial red-teaming, and prompt injection mitigation to prevent hallucination, data leakage, and unauthorized tool execution within corporate cloud environments. Build the Benchmarking Infrastructure: Construct robust evaluation datasets, golden regression sets, and synth-data pipelines that mimic real-world property manager, tenant, and broker behaviors. Monitor System Drift & Observability: Monitor live, production-grade agent workflows for behavioral drift, latency regressions, cost inefficiencies, and token utilization over time. Build Automated Evaluation Infrastructure: Design, develop, and maintain scalable evaluation frameworks and pipelines using Python, SQL, APIs, and CI/CD to execute hundreds or thousands of evaluation cases across models, prompts, RAG pipelines, and AI agents. Develop Evaluation Metrics and Graders: Implement automated grading approaches using deterministic validation, reference-based evaluation, human-defined rubrics, LLM-as-a-Judge, and human-in-the-loop evaluation. Establish scoring criteria and quality thresholds appropriate to each AI use case. Required Qualifications 5 to 7 years of experience in software engineering, quality engineering, test automation, AI engineering, machine learning, or a related technical role. Hands-on experience evaluating or testing LLM, NLP, machine learning, or AI applications. Strong Python skills and experience building automated test or evaluation workflows. Experience creating and maintaining golden datasets, test datasets, or benchmark datasets. Experience with adversarial testing. Familiarity with AI Safety, security & Privacy, and Responsible AI evaluation. Deep structural expertise in LangChain, LangGraph (including sub-graphs and conditional routing) and LangSmith. Hands-on experience using LangSmith for tracing, debugging, experiments, datasets, and evaluations. Strong GitHub experience, including repositories, branching, pull requests, code reviews, and CI/CD workflows. Experience using Claude Code or similar AI-assisted development tools. Experience with AWS and Amazon Bedrock. Experience with REST APIs, JSON, SQL, and modern application architectures. Strong understanding of LLMs, prompt engineering, embeddings, vector search, RAG, and AI agents. Experience with automated testing, regression testing, integration testing, and end-to-end testing. Hands-on experience evaluating LangGraph-based or multi-agent applications at the node and state-transition level. Strong analytical, problem-solving, and communication skills. Ability to work independently and take ownership of ambiguous technical problems. Preferred Qualifications Experience with Ragas, DeepEval, Arize Phoenix, TruLens, Promptfoo, or similar evaluation tools. Experience with LLM-as-a-judge and human-in-the-loop evaluation approaches. Experience with Bedrock Agents, Knowledge Bases for Amazon Bedrock, Lambda, S3, or CloudWatch. Experience integrating AI evaluations into GitHub Actions.