AI Engineer
Summary
Designs and ships agentic AI systems — multi-agent workflows, RAG pipelines, and enterprise integrations — for a bilingual enterprise knowledge platform. Core work spans Python, LLM orchestration frameworks like LangGraph/LangChain, vector and lexical retrieval, and background job infrastructure.
We are seeking an Expert AI Engineer to own the design, architecture, and deployment of agentic AI systems for a bilingual enterprise knowledge platform built on large language models. You will define the technical direction for autonomous, multi-agent, and multi-step workflows—incorporating planning, reasoning, memory, and tool use—on top of production-grade Retrieval-Augmented Generation (RAG) pipelines, along with the workflow, integration, and background processing layers that support them. This role carries deep end-to-end ownership of architecture, quality, and performance.
Responsibilities
Architect agentic systems incorporating query planning, reasoning, memory, tool use, and multi-step task execution
Define orchestration patterns that coordinate agents, retrieval, tools, and LLM calls into reliable, observable autonomous pipelines
Design robust single- and multi-agent architectures with state management, control flow, error recovery, and guardrails
Architect the workflow layer: composable, versioned workflows combining deterministic steps, agentic branches, conditional routing, and human-in-the-loop checkpoints, with durable state, checkpointing, and retries
Design the connector and integration layer for enterprise content sources and APIs, covering authentication, incremental sync, content normalization, and permission-aware retrieval
Own the background processing layer, including scheduled ingestion, index and embedding refresh, job queuing and concurrency control, failure recovery, and content-freshness monitoring
Own and optimize the underlying RAG layer spanning chunking, embedding, retrieval, reranking, and grounded generation
Implement and tune dense, sparse, hybrid, and metadata-based retrieval using vector databases and BM25
Establish prompting, grounding, and verification strategies to ensure responses remain accurate and citation-backed
Define evaluation frameworks and quality gates; drive continuous improvement across task success, relevance, latency, and reliability
Requirements
Deep, demonstrated expertise in agentic AI and LLM-based applications, with a track record of production systems
Expert Python proficiency and a strong software engineering foundation, including production deployment and system design
Proven experience architecting agentic systems with planning, tool use, orchestration, and multi-agent coordination
Deep hands-on experience with agent frameworks such as LangGraph, LangChain, LlamaIndex, AutoGen, or CrewAI
Strong command of agent state management, tool/function calling, and standards such as the Model Context Protocol (MCP)
Demonstrated experience designing workflow orchestration systems, including DAG or state-machine execution, durable state, retry semantics, and idempotent step design
Hands-on experience building connectors and integrations against enterprise systems and third-party APIs, including OAuth and service-account authentication, rate limiting, and permission-aware synchronization
Production experience with background job and scheduling infrastructure (Celery, Airflow, Prefect, Temporal, or equivalent), including scheduled ingestion and failure recovery
Hands-on experience with vector databases (Milvus, FAISS, Qdrant, pgvector) and lexical search (BM25, Elasticsearch)
Expert command of retrieval techniques, including dense, sparse, hybrid, and filtered search, plus reranking
Advanced prompt engineering and grounding techniques for large language models
Track record of deploying, scaling, and optimizing production-ready agentic AI pipelines
Strong grasp of RAG and agent evaluation methodologies, with the ability to define quality standards