AI Engineer — Agentic Systems

Cynet Systems•Austin, TX

About The Position

The AI Engineer designs, builds, and operates agentic AI systems: LLM-powered agents that reason over enterprise data, call tools, and execute multi-step workflows. This hands-on engineering role turns ambiguous technical and operational problems into agent workflows and production services. The engineer works within an established platform and architecture, with senior engineers available for design review, and owns components from prototype through production. The role focuses on agent orchestration and multi-agent workflows; retrieval-augmented generation and enterprise knowledge systems; production Python services and APIs; deployment, observability, and LLMOps; and evaluation and quality. The engineer also partners with domain experts, data engineers, and product stakeholders to automate manual workflows, deliver user-adopted interfaces, document designs and runbooks, and participate in code and design reviews.

Requirements

3–5 years of professional software, ML, or data engineering experience, including at least 1 year building LLM-based or GenAI applications.
Strong Python engineering skills, including production code, testing, packaging, asynchronous patterns, and API development with FastAPI or similar.
Hands-on experience with at least one LLM orchestration framework—LangChain, LangGraph, LlamaIndex, Semantic Kernel, or equivalent—in a shipped application.
Demonstrated experience building RAG systems, including embeddings, vector stores such as FAISS, pgvector, Pinecone, or Chroma, and retrieval-quality tuning.
Practical command of prompt engineering, structured output, function/tool calling, and context-window management.
Working knowledge of SQL and cloud data platforms such as Snowflake, Databricks, or equivalent, and comfort querying enterprise schemas.
Experience containerizing and deploying services with Docker and shipping through CI/CD on a major cloud platform: Azure, AWS, or GCP.
Solid grounding in ML fundamentals and familiarity with PyTorch, TensorFlow, or HuggingFace Transformers.
Strong verbal and written communication skills, and judgment to escalate when a problem is under-specified rather than guessing.

Nice To Haves

Experience building or consuming Model Context Protocol (MCP) servers, or comparable agent tool-integration layers.
Production experience with multi-agent or agentic workflow systems, including human-in-the-loop checkpoints and approval gates.
Familiarity with LLM observability and evaluation tools such as LangSmith, Langfuse, Weights & Biases, or MLflow.
Experience with workflow orchestration tools such as Airflow, Dagster, or Prefect, or analytics engineering with DBT.
Exposure to model fine-tuning or parameter-efficient adaptation (LoRA/PEFT), and knowledge of when it is or is not appropriate.
Background in a technical engineering domain such as semiconductor, hardware validation, or manufacturing, where agents operate over specialized data.
Experience with AI governance, responsible-AI review, or enterprise security requirements for LLM deployments.

Responsibilities

Build and ship LLM-powered agents that plan, call tools, and complete multi-step tasks using frameworks such as LangChain, LangGraph, or equivalent libraries.
Implement multi-agent workflows with planner, executor, reviewer, and specialist roles, including state handling, hand-off logic, retries, and graceful failure paths.
Design and expose agent tools through function-calling and MCP servers, wrapping internal APIs, databases, and enterprise systems as safe, well-scoped capabilities.
Build and operationalize RAG pipelines, including chunking, embedding, vector and hybrid search, re-ranking, and context assembly.
Integrate structured, semi-structured, and unstructured sources such as SharePoint, Confluence, Jira, REST APIs, Excel, and SQL warehouses into retrieval and agent context layers.
Improve answer grounding and reduce hallucination through retrieval, prompt design, and context-management improvements.
Develop production Python services and APIs for agent and model inference using FastAPI or similar, containerize them with Docker, and deploy through standard CI/CD pipelines.
Instrument agents with tracing, token and cost accounting, latency monitoring, and structured logging to make production behavior observable and debuggable.
Apply MLOps and LLMOps practices, including version control for prompts and configurations, automated regression testing, and safe rollout of model or prompt changes.
Build evaluation harnesses for agent workflows, including golden datasets, task-level success metrics, LLM-as-judge scoring, and release-gating regression suites.
Measure and tune accuracy, latency, and cost, making explicit trade-offs based on the use case.
Partner with domain experts, data engineers, and product stakeholders to turn manual workflows into reliable automated agents.
Deliver user-adopted interfaces such as Streamlit apps, chat interfaces, dashboards, and embedded integrations.
Write design notes and runbooks, and participate in code and design reviews.