Job Description
About the Role :
At BeeHyv, we build AI-powered products and agent systems for enterprise clients across healthcare, insurance, and government. This is a hands-on engineering role for someone who already builds full stack web applications and has started shipping real features on top of LLMs not just demos.
You will work close to the problem : turning client requirements into working software, deciding how (and whether) to apply an LLM, and shipping reliable, observable, maintainable systems. You will pair with senior engineers and forward-deployed teams, so you will grow quickly on both the product and AI sides.
What You'll Work On
- Build and ship full stack features APIs, data models, and front-end for client-facing AI products.
- Implement RAG pipelines over enterprise documents and databases : ingestion, chunking, embeddings, and retrieval.
- Build agentic workflows and tool-calling features using Claude and orchestration frameworks such as LangGraph.
- Add the unglamorous-but-essential parts : guardrails, evals, logging, tracing, fallbacks, and cost/latency monitoring.
- Integrate with client systems and third-party APIs, and help take features from prototype to production.
Core Stack
- Backend : Python (FastAPI) or Java (Spring Boot) strong depth in one. Solid REST API design; comfort with async work, background jobs, and queues.
- Frontend : A modern framework React, Angular, or Vue with TypeScript. Clean, responsive UI; sensible component structure and state management.
- Data : PostgreSQL (including pgvector) and Redis. Schema design and query tuning.
- AI layer : Anthropic / Claude APIs, embeddings and vector search, plus an orchestration framework (LangGraph or LangChain).
- Cloud & DevOps : AWS or Azure, Docker, and a working understanding of CI/CD and deployment.
Required
What We're Looking For :
- 3-5 years building and shipping production web applications, end to end.
- Strong in one backend Python or Java/Spring Boot and one modern front-end framework (React, Angular, or Vue with TypeScript).
- 1-2 years of hands-on LLM application work - you've built at least one real RAG or tool calling feature that ran in front of users.
- Working knowledge of prompt design, embeddings, vector search, and chunking trade offs.
- Hands-on multi-agent orchestration - you've built agentic/tool-calling workflows with a framework such as LangGraph, CrewAI, AutoGen, or Semantic Kernel.
- Practical AI observability - tracing, evals, and monitoring of LLM features (e.g., Langfuse, Grafana, or equivalent), including cost and latency.
- Comfort with SQL databases (Postgres preferred) and basic cloud deployment (AWS or Azure).
- Clear written and verbal communication - you can explain trade-offs to engineers and to clients.
Nice To Have
- Cloud AI services : AWS Bedrock, Azure AI Foundry, or Vertex AI.
- Experience in regulated domains - healthcare, insurance, or government.
- Node.js, alternative vector stores (Pinecone, Weaviate, Qdrant, Chroma, FAISS), or local/open-source model deployment.
How We Evaluate
Our process is designed to respect your time :
- Intro conversation - your background and what you've built.
- Technical discussion - a walk-through of a real AI feature you've shipped, plus a focused full stack + LLM exercise.
- System & design conversation - how you'd design a RAG or agent feature, including guardrails and trade-offs.
- Final conversation with the engineering lead.
(ref:hirist.tech)
