Mid level Agentic AI Engineer

About the Role

------------------

You will design, build, and operate production-grade multi-agent systems for commercial and federal customers. The work spans the LangChain ecosystem, reliable agent loops, advanced retrieval architectures, and ontologies that ground agent reasoning.

You will also connect agents to enterprise tools and to each other through MCP and A2A. This role suits an engineer who thinks in systems: how agents reason, retrieve, act, recover from failure, and get evaluated and observed in production.

Key Responsibilities

------------------------

Agent development and orchestration

• Build single- and multi-agent applications with LangChain and LangGraph, including stateful graphs, supervisor/worker and hierarchical patterns, and human-in-the-loop checkpoints.
• Engineer robust agentic loops (plan, act, observe, reflect) with clear termination conditions, retry and self-correction logic, guardrails, and cost and latency budgets.
• Apply loop engineering to tune iteration depth, tool-call strategy, memory handling, and state persistence so agents behave predictably.
• Use LangSmith for tracing, debugging, prompt and dataset management, and automated evaluation; maintain regression suites for agent behavior.

Retrieval-Augmented Generation (RAG)

• Build and optimize Agentic RAG (agent-driven query planning and iterative retrieval), Adaptive RAG (routing by query complexity), Corrective and Self-RAG, GraphRAG (knowledge-graph retrieval and reasoning), and Cache-Augmented Generation with semantic caching.
• Design chunking, embedding, hybrid search, re-ranking, and query-rewriting strategies.
• Measure retrieval quality with faithfulness, context precision and recall, and answer relevance.

Ontologies, context, and skills

• Develop domain ontologies and knowledge graphs (OWL/RDF or property-graph models) that ground agent reasoning and power GraphRAG.
• Practice context engineering: decide how instructions, memory, retrieved knowledge, and tool results are assembled in the context window.
• Build reusable agent skills and tool libraries with clear interfaces, documentation, and versioning.

Interoperability protocols

• Build and integrate Model Context Protocol (MCP) servers and clients that expose enterprise data, APIs, and tools to agents securely.
• Implement Agent-to-Agent (A2A) patterns for agent discovery, capability negotiation, task delegation, and cross-vendor communication.

Engineering and delivery

• Write production-quality Python (TypeScript optional) with tests, CI/CD, and secure coding practices.
• Take agents from prototype to production with product, data, and platform teams.
• Contribute to architecture decisions, code reviews, and internal standards for agentic systems.

Required Qualifications

---------------------------

• Bachelor's degree in Computer Science, Engineering, or a related field, or equivalent experience.
• 10 years of software engineering, with at least 4 years building LLM-powered applications.
• Hands-on production experience with LangChain, LangGraph, and LangSmith.
• Demonstrated experience designing agentic loops, tool calling, memory, and multi-agent orchestration.
• Practical experience implementing several RAG variants (Agentic, Adaptive, GraphRAG, Cache RAG/CAG) and evaluating their quality.
• Experience designing ontologies or knowledge graphs and applying them to retrieval or reasoning.
• Strong context engineering and prompt design for complex agent workflows.
• Experience building or integrating MCP servers and clients, and working knowledge of the A2A protocol.
• Proficiency in Python, REST/gRPC APIs, and asynchronous programming.
• Solid grasp of LLM failure modes (hallucination, prompt injection, context limits) and their mitigations.

Trillion is an equal opportunity employer. Applicants are considered for positions without regard to race, religion, gender, native origin, age, disability, or any other category protected by applicable federal, state, or local law.