AI / Agent Engineer (Principal / Lead)

About Aura Cloud Aura Cloud is a class leading cloud platform for financial institutions to launch and run their business at light-speed. Our platform is at the center of our customer's digitization path and business critical. We provide 24/7 platforms for critical core financial applications that require a very high degree of quality and stability of the platform. We are fast expanding into other markets with continued focus in the Nordics We seek an experienced AI / Agent Engineer (Principal / Lead) to join the Platform Engineering Team at our Bengaluru office, to shape and drive the development of Aura Cloud's AI-powered solutions and agentic systems across our core banking and financial platforms. About AuraCrew AuraCrew is an AI-native platform for building, governing and deploying AI agents and agentic workflows (“SmartFlows”) for banks and insurers — with model-risk and regulatory governance built into the product. The Role This is the differentiating core — the agents, the orchestration, and the evaluation / guardrail / oversight machinery that lets a bank trust an AI decision. What You'll Own AuraCrew’s AI core on AWS Bedrock / AgentCore: agent authoring and runtime, RAG / knowledge, and the AI Shield / Eval Center / oversight layers. What You'll Do Design and productionize agentic systems: multi-step orchestration, tool-calling, structured outputs, agent memory — deployed to AgentCore Runtime. Build the evaluation and guardrail layer that is the product’s reason to exist: source-grounding, omission / hallucination detection, eval suites, red-teaming, trust scoring, model inventory. Own model governance in-product: model bindings across providers, prompt-injection and least-privilege defenses, oversight rules, token / cost accounting. Push RAG / retrieval quality: hybrid search, reranking, multi-tenant knowledge, data-model-aware retrieval. Set the standard for how AI systems are built, evaluated and shipped here. What You'll Bring 8+ years total; 3+ years building and productionizing LLM / agent systems. Demonstrated production work with: agent orchestration (LangGraph / LlamaIndex / Agents SDKs / MCP or equivalent), evals & guardrails (RAGAS / DeepEval / Langfuse or your own), and RAG. AWS Bedrock (or equivalent managed LLM infra); MLOps; strong Python; comfortable in TypeScript. An AI-security mindset: prompt-injection, least-privilege identity, data-leakage controls. Bonus AgentCore Runtime; DSPy / prompt-optimization; voice / realtime; fine-tuning & serving open-weight models; published or open-source agent work. Not The Right Fit If Your agent experience is demos and notebooks, or you’ve never had to prove an AI system’s outputs were safe and correct. How We Hire One bar, tested the same way: a deep dive on real work you’ve shipped — architecture, trade-offs, and what was actually yours — then a paired working session on a live AuraCrew problem in your area. We optimise for depth, judgement and ownership, not breadth of buzzwords. Stack NestJS · PostgreSQL · React / TypeScript · AWS (Bedrock, AgentCore Runtime, Lambda, S3, IAM / KMS / Secrets Manager, CloudFormation) · Python agent runtime · OpenTelemetry