Agentic Orchestration: Design and implement agentic workflows using frameworks like LangGraph, CrewAI, or Semantic Kernel.
Cloud Solution Design: Architect and deploy production-grade applications on at least one major AI platform:
Azure AI Foundry: Utilizing AI Agent Service and Semantic Kernel for enterprise integration.
AWS Bedrock: Building managed RAG pipelines and autonomous agents using Bedrock Agents.
Google AI Studio / Vertex AI: Implementing long-context solutions with Gemini and the Agent Development Kit (ADK).
System Integration: Oversee the integration of AI models with existing IT infrastructure, microservices, and vector databases (Pinecone, Weaviate, etc.).
Governance & Safety: Implement "Responsible AI" guardrails, including prompt injection protection, PII filtering, and output validation.
Fast-Paced Innovation: Rapidly prototype and evaluate new releases (e.g., O1 reasoning models, new multimodal capabilities) to determine business viability.
Key Skills Required:
Model Orchestration: Deep understanding of Chain-of-Thought (CoT) reasoning and ReAct (Reason+Act) patterns.
Cloud Proficiency: Hands-on experience with managed AI services (Azure AI Studio, Bedrock, or Vertex AI).
Tool Use (Function Calling): Expertise in teaching models how to use external tools, APIs, and databases accurately.
Development Mastery: Expert Python or TypeScript skills, focusing on asynchronous patterns and API design.
Architecture Patterns: Experience with Event-Driven Architecture and Microservices.
Added Advantages:
Prior ML App History: Experience building and scaling machine learning-related apps (e.g., recommendation engines, predictive analytics) before the LLM era.
Advanced Tooling: Familiarity with high-productivity AI developer tools is a major plus:
Terminal/CLI Agents: Experience with Gemini CLI or Claude Code for autonomous repository management.
IDE Integration: Advanced usage of GitHub Copilot (including Agentic Chat) or Cursor to accelerate the SDLC.
Evaluation (LLM-as-a-Judge): Experience building automated evaluation frameworks to measure agent performance and reliability.
Rapid Learning: A proven ability to transition from "knowing nothing" to "building a PoC" with a new framework in under 72 hours.
