Job Title: Junior AI/ML Developer Location: Fremont, CA Duration: Fulltime
Key Responsibilities
Agent Development: Assist in building and deploying intelligent agents using Vertex AI Agent Builder and Agent Engine.
Protocol Implementation: Write and maintain MCP Servers to connect LLMs to external tools, databases, and Salesforce environments.
A2A Orchestration: Develop standardized Agent Cards and implement messaging logic that allows different AI agents to collaborate on complex tasks.
Cloud Infrastructure: Set up and manage basic AI workflows on Google Cloud Platform, including Cloud Run deployments and Pub/Sub messaging for agents.
Testing & Tuning: Conduct rigorous testing of agentic "reasoning" loops and perform prompt engineering to optimize model performance.
Technical Requirements
Education: Bachelor s or Master s in Computer Science, Data Science, or a related quantitative field.
Core Language: Strong proficiency in Python (AsyncIO, FastAPI/Flask).
AI Fundamentals: Understanding of LLMs (Gemini, GPT), RAG architectures, and vector databases.
Cloud Exposure: Familiarity with AWS, Azure, Google Cloud Platform.
Version Control: Proficiency with Git and collaborative development workflows.
Hands-on experience with LangChain, CrewAI, or LangGraph.
Familiarity with JSON-RPC (the backbone of MCP/A2A).
Active GitHub profile or participation in AI hackathons.
Category Must-Have (Level 1-2)
Languages Python (Advanced: AsyncIO, Typing, Pydantic), SQL, Node.js, TypeScript (useful for MCP servers)
AI Foundations LLM APIs (Gemini, GPT-4), Prompt Engineering, RAG, Agentic Workflows, Chain-of-Thought reasoning
Protocols JSON-RPC 2.0, REST APIs, WebSockets, MCP, A2A Architecture
Cloud Google Cloud Platform, Azure, AWS
Frameworks LangChain, FastAPI, Flask, LangGraph, CrewAI, LlamaIndex
Data & Vectors Vector DBs (Pinecone, Chroma, or Milvus), Google Vector Search, Firestore
DevOps/Tools Git/GitHub, Docker, Basic CI/CD, MLOps basics, Weights & Biases, LangSmith
Architecture Microservices basics, JSON/XML parsing, Agent Cards, Tool-calling definitions