Heu.ai

Overview

We are looking for a hands-on AI / LLM Engineer to design, build, and deploy intelligent system architectures and agentic workflows. In this role, you will go beyond simple wrapper scripts to build production-grade Retrieval-Augmented Generation (RAG) pipelines, complex multi-agent orchestrations, and robust backend APIs.

About Heu.ai

Heu.ai is an AI platform company focused on helping enterprises enhance their business and solve complex problems through AI-driven solutions.

Requirements & Eligibility

• Strong proficiency in Python with an emphasis on async programming, clean architecture, and type safety.
• Demonstrated hands-on experience with LangGraph, LangChain, or custom agentic execution systems.
• Solid background in building and maintaining production APIs via FastAPI or modern Web frameworks, including authentication, rate-limiting, and error-handling patterns.
• Proven experience implementing end-to-end RAG workflows with vector stores (Pinecone, Qdrant, PGVector, etc.) and embedding models.
• Mastery of structured outputs (JSON schema, Pydantic), chain-of-thought prompting, and defensive prompt engineering against edge cases and injection.
• Experience integrating third-party APIs, webhooks, and custom tools into LLM workflows.
• Experience with fine-tuning open-source models (e.g., Llama, Mistral) or local LLM deployments (Ollama, vLLM).

Key Responsibilities

• Design and deploy stateful, multi-agent orchestrations and dynamic decision-making graphs using LangGraph.
• Build, evaluate, and optimize hybrid retrieval strategies (vector search, keyword/BM25, metadata filtering, reranking) and chunking techniques for high-precision context retrieval.
• Architect, test, and iterate on structured prompts, system instructions, and dynamic context windows.
• Establish evaluation frameworks to monitor quality and hallucination rates.
• Develop scalable, asynchronous REST/gRPC APIs and microservices using Python frameworks (FastAPI, Flask) to integrate AI capabilities into production applications.
• Manage and optimize vector databases for efficient embedding indexing and search performance.
• Monitor latency, token usage, rate limits, and cost across LLM providers.
• Implement caching strategies and streaming outputs for low-latency end-user experience.