Key Responsibilities:
Agent Architecture & Development:
- Design and implement autonomous agent systems using frameworks like Hermes Agent.
- Build multi-agent collaboration patterns (e.g., orchestrator-workers, debate, hierarchical swarms).
- Implement agentic memory systems (short-term, long-term, and episodic memory) using vector databases and semantic caching.
Reasoning & Planning:
- Integrate advanced reasoning techniques: ReAct, Chain-of-Thought (CoT), Tree-of-Thoughts (ToT), and Plan-and-Solve.
- Develop agents capable of dynamic planning, error recovery, and replanning based on environmental feedback.
- Implement tool use (function calling) and API grounding for actions like database queries, API calls, RAG retrieval, and UI automation.
Production & Evaluation:
- Build robust evaluation frameworks (agentic eval) to test for task completion, efficiency, and safety—not just lexical similarity.
- Instrument agents with tracing, observability, and logging (e.g., LangSmith, Arize, Weights & Biases).
- Optimize for latency, cost (token usage), and reliability in production.
Integration & Tooling:
- Connect agents to internal and external systems: CRMs, databases, Slack, browsers, REST APIs, and code interpreters.
- Develop custom tools and sandboxed environments for agents to execute code or shell commands safely.
Required Qualifications:
Technical Skills:
- Programming: Expert in Python.
- Strong understanding of prompt engineering, few-shot learning, and structured output generation (JSON mode, grammars).
- Reasoning Patterns: Proven experience implementing agentic patterns (ReAct, Reflexion, Toolformer) in production or complex prototypes.
- Memory & Retrieval: Experience with vector databases (Pinecone, Weaviate, Qdrant) and RAG optimization (hybrid search, reranking).
- Orchestration: Familiarity with workflow engines (Temporal, Prefect, Airflow) for human-in-the-loop and durable execution.
- Observability: Experience monitoring LLM applications (prompt traces, token usage, drift).
- Model Context Protocol: Built agents that use MCP for multi-step research, code analysis, or data engineering tasks.
- Agentic Framework: Practical experience with Hermes Agent.
Education & Experience:
- Bachelor’s degree in Computer Science, Software Engineering, AI, or related discipline.
- 3 years in software engineering / ML engineering.
- Experience building production-grade agentic systems (not just demos or chatbots).
- Strong understanding of LLM limitations: hallucinations, jailbreaks, prompt injection, and failure modes.
- Good understanding of MCP discovery patterns and context negotiation.
- Strong knowledge of context management in LLM applications: prompt caching, sliding window, semantic retrieval, MCP resource lifecycle.
