Key Responsibility
• Design, develop, and maintain end-to-end AI/ML solutions from experimentation to scalable production systems.
• Collaborate with data scientists to productionize models (ML pipelines, APIs, containers, model registries).
• Build reusable full-stack components to integrate AI into front-end and back-end banking systems.
• Ensure model versioning, CI/CD, monitoring, and automated retraining pipelines follow governance standards.
• Optimize ML workloads and infrastructure using scalable data platforms (e.g., Spark, Kubernetes, cloud).
• Troubleshoot deployment issues and implement observability tools to monitor AI performance in production.
• Work closely with product teams to validate business requirements and deliver on use case objectives.
Front-End & User Experience
• Build React/Next.js dashboards that visualise multi-step agent plans, tool calls and audit trails, giving users transparency into autonomous decision-making.
• Implement real-time SSE/WebSocket channels to stream agent chain-of-thought with redaction safeguards for sensitive prompts.
Back-End & Agentic Middleware
• Develop micro-services (FastAPI / NestJS / Spring Boot) acting as MCP servers that expose bank-internal tools (payment rails, credit-risk engines) to agents.
• Integrate A2A bridges so agents authored in different stacks (Python, Go, JS) interoperate across cloud and on-prem clusters.
• Orchestrate agentic RAG pipelines—Weaviate/Chroma vector search + task-planning agents—for complex knowledge queries.
DevOps
• Automate CI/CD that unit-tests agent tool chains, runs synthetic evals and pushes blue-green releases behind feature flags.
Security, Compliance & Governance
• Embed runtime policy guards: tool-call whitelisting, prompt-template validation, and MCP context-size accounting to prevent data exfiltration.
• Capture lineage: every agent action → tool → SQL, stored in immutable audit logs for regulators and model-risk teams.
Collaboration & Leadership
• Mentor engineers on agentic design patterns (planner-executor, self-reflection loops).
• Publish internal white-papers comparing MCP vs A2A capabilities and migration paths.
EDUCATION
• Bachelor’s or Master’s degree in Computer Science, Software Engineering, Data Science, or related field.
TRAINING
• Open-source contributions to agentic frameworks (e.g., LangChain, Llamaindex, Autogen, CrewAI).
• Experience with agentic RAG and vector-DBs (e.g., Chroma, pgvector).
• Familiarity with AI model governance in finance.
A plus point:
• Operational experience on Nvidia comparable GPU clusters (≥1 PFLOPS)
• Familiarity with Hyperscale GPU economics, Spot vs. Savings-Plan optimisation, and GPU orchestration platforms
• Track record fine-tuning 10 B+ parameter LLMs with LoRA/QLoRA, releasing checkpoints on Hugging Face
LICENSES
• Google Cloud Professional ML Engineer or AWS Certified Machine Learning – Specialty
• Microsoft Azure AI Engineer Associate
• Certified Kubernetes Administrator (CKA) or Docker Certified Associate
• TensorFlow Developer Certificate (optional)
• Familiarity with Responsible AI, model interpretability, or compliance tools is a plus
MEMBERSHIP
• Participation in AI communities such as Data Science Society, PyData, or local ML meetups
CERTIFICATIONS
ML engineer, software engineer, AI/ML
LANGUAGES
English ; Chinese