Role Overview
We are hiring AI Engineers to build, ship and operate Generative AI and Agentic AI capabilities for Enterprise Banking customers across Azure, AWS and GCP. You will work in a tight-knit pod led by an AI Lead, partnering with data engineers, full-stack developers, designers and banking domain SMEs to convert real business problems into reliable, secure, well-governed AI systems.
This is a deeply hands-on engineering role. You will write production code, design prompts and chains, build and evaluate RAG pipelines, develop multi-agent workflows, and own services through to observability and incident response. You will care about correctness, latency, cost, safety, and the regulatory posture that banking demands.
Key Responsibilities
1. Build & Ship GenAI / Agentic Features
- Develop and deploy RAG pipelines, agent workflows, copilots, and AI-powered microservices on Azure, AWS or GCP — typically across more than one cloud per engagement.
- Implement chunking, embedding, indexing, hybrid retrieval and re-ranking strategies tuned to banking corpora (policy documents, regulatory text, contracts, customer correspondence, code, structured data).
- Design and implement tool-use, function-calling and structured-output patterns; orchestrate multi-step agents with planning, reflection, memory and human-in-the-loop checkpoints.
- Integrate with banking systems of record (core banking, CRM, ECM/document repositories, fraud platforms, ticketing systems) via secure APIs and event streams.
2. Quality, Evaluation & Safety
- Build and maintain golden datasets and evaluation harnesses; instrument pipelines with metrics for groundedness, faithfulness, latency-per-token, cost-per-call and user-feedback signals.
- Implement guardrails — input filtering, output validation, schema enforcement, PII redaction, prompt-injection defenses and jailbreak detection.
- Participate in red-teaming exercises and remediate findings; contribute to model cards and decision logs for audit and MRM reviews.
3. MLOps / LLMOps & Reliability
- Containerize services, write infrastructure-as-code, and ship through CI/CD pipelines including evaluation gates for prompts, chains and agents.
- Set up observability for AI systems — distributed tracing, prompt/response logging (with redaction), token-level cost dashboards, and alerting on quality regressions.
- Own production support for your services on a rotating basis; conduct blameless post-mortems and drive permanent fixes.
4. Collaboration & Continuous Learning
- Partner with product managers, designers and banking SMEs to shape requirements and demonstrate progress through frequent, working increments.
- Contribute to internal accelerators, reusable components, and the AI engineering knowledge base; mentor junior engineers and interns.
- Stay current with the rapidly evolving GenAI/Agentic ecosystem — models, frameworks, evaluation techniques, and security patterns — and bring relevant innovations into the practice.
Required Skills & Qualifications
Core Engineering
- Strong programming skills in Python (typing, async, packaging, testing); secondary proficiency in TypeScript/Node.js, Java or Go is a plus.
- Solid grasp of data structures, algorithms, system design, REST/gRPC API design, and asynchronous/event-driven patterns.
- Comfortable with FastAPI, SQLAlchemy (or equivalent), Pydantic, Redis, Celery / Kafka / Pub/Sub / SQS, PostgreSQL, and at least one NoSQL store.
- Hands-on with Docker, Kubernetes, Git workflows, and at least one CI/CD platform (GitHub Actions, Azure DevOps, GitLab CI, Cloud Build).
GenAI & Agentic AI
- Practical experience building RAG systems end-to-end: ingestion, chunking, embeddings, vector search, retrieval, re-ranking, and grounded generation.
- Experience with at least one agentic framework: LangGraph, LangChain, LlamaIndex, Semantic Kernel, CrewAI, AutoGen, Google ADK, or Pydantic AI.
- Comfortable working with frontier models (OpenAI GPT, Anthropic Claude, Google Gemini) and open-weight models (Llama, Mistral, Qwen, Phi, DeepSeek) via managed and self-hosted inference.
- Skilled in prompt engineering, structured outputs (JSON-mode, tool calling), function/tool design, and context-window management.
- Familiarity with evaluation tools — Ragas, DeepEval, Promptfoo, TruLens, LangSmith, Arize Phoenix — and ability to design custom evaluation harnesses for domain-specific tasks.
- Working knowledge of fine-tuning techniques (SFT, LoRA/QLoRA, DPO) and when to use them versus prompting and RAG.
Cloud Platforms (Multi-Cloud Required)
Engineers are expected to be strong on at least one of the three platforms below and working-level on a second, with willingness to grow into the third:
- Azure: Azure OpenAI / AI Foundry, AI Search, AI Content Safety, Document Intelligence, Functions, AKS, Cosmos DB, Key Vault, Entra ID.
- AWS: Bedrock (Knowledge Bases, Agents, Guardrails), SageMaker, OpenSearch, Lambda, ECS/EKS, API Gateway, KMS, IAM.
- GCP: Vertex AI (incl. Agent Builder), Gemini API, BigQuery (incl. vector search), Cloud Run, GKE, Cloud KMS.
- Comfortable with private networking, identity, secrets management and cost-aware design across at least one cloud, and motivated to extend the same patterns to the others.
Banking Awareness (Welcome but Trainable)
- Awareness of common banking domains and data types — accounts, transactions, KYC documents, credit memos, policies, regulatory filings, customer communications.
- Sensitivity to data privacy, data residency, PCI/PII handling, and the importance of auditability.
- Willingness to learn applicable regulations (RBI, MAS, FCA, OCC, GDPR, DPDP) and adapt designs accordingly.
Soft Skills
- Clear written communication — design docs, PR descriptions, runbooks, and customer-facing summaries.
- Pragmatism in balancing innovation against reliability, security and cost.
- Curiosity, ownership, and ability to operate with limited supervision in ambiguous problem spaces.
Preferred / Nice-to-Have
- Cloud certifications (Azure AI Engineer Associate, AWS ML Specialty, Google Professional ML Engineer).
- Experience with knowledge graphs and graph-RAG (Neo4j, Neptune, Spanner Graph), or with FIBO and other banking ontologies.
- Exposure to model serving frameworks — vLLM, TGI, NVIDIA Triton/NIM — and to on-prem or sovereign deployments.
- Open-source contributions, public technical writing, or competition placements (Kaggle, hackathons).
- Frontend or full-stack familiarity (React, Angular, Vue) for building copilot UIs.
