Senior GenAI Engineer
你将负责从想法到开发和部署，并持续改进的 LLM 驱动功能，重点关注答案质量，性能，可观察性和成本效益。你将与现代 LLC 技术合作，包括兰格拉夫，RAG,工具调用，Azure OpenAI,双子座和克劳德，并且对人工智能驱动产品如何构建和运营在生产中产生真正影响。责任：· 建立 LLM 驱動功能终结至终结。设计和实施使用 LangGraph 的代理流程，检索和工具调用 - 然后将它们作为 FastAPI 服务提供了流媒体，持久性和适当测试。· 拥有答案质量。构建评估数据集，回归套件和 LLM-as-judge 检查，以便我们知道提示或模型更改是否使事情变得更好之前它达到用户。· 将用户问题转化为针对我们的搜索平台有效查询，编制多步骤研究循环，并塑造背景模式原因。当一个回答是错误时，确定检索，查询或提示是否有错误 - 并修复正确的问题。· 调整生产。工具流动与追踪 (Langfuse),从真实痕迹中调查错误的答案，管理延迟，代币和成本预算 - 包括在 AI 门口后面的模型尺寸和家庭之间路由。要求 ·构建和运营后端服务的经验 - API,异步，测试 ·将 LLM 功能带到生产中并使其运行中的实践经验 - 不仅是原型程序 ·代理 /编排框架 - LangGraph 理想 •实际 RAG 经验 · 在生产中调整 LLM 系统的经验-追踪，评估，成本和延迟 ·在云计算运行的服务经验 （我们是在 Azure 上）· 强大的解决问题技能，分析思维和技术决策 · 英语流利，主动沟通者以及合作团队玩家 · 开放思想，创造力，并有动机推进人工智能和自动化界限：· Azure OpenAI, AI 搜索，应用服务 · 基础设施作为代码 (Bicep) · 导师或科技领导经验：· Python · FastAPI · LangGraph · Azure Open AI, Google Gemini,人类 Claude · FAISS · PostgreSQL · Langfuse · Azure App Insigh
查看英文原文
This is a remote position.
We are looking for a Senior AI Engineerto join a team building and operating a production-grade LLM system used by real users. This is not a proof-of-concept project. You will be responsible for taking LLM-powered features from idea through development and deployment to continuous improvement, with a strong focus on answer quality, performance, observability and cost efficiency. You will work with modern LLM technologies, including LangGraph, RAG, tool calling, Azure OpenAI, Gemini and Claude, and have a real impact on how AI-powered products are built and operated in production.
Responsibilities:
· Build LLM-powered features end to end. Design and implement agentic flows, retrieval, and tool calling using LangGraph - then ship them as FastAPI services with streaming, persistence and proper tests.
- Own answer quality. Build evaluation datasets, regression suites and LLM-as-judge checks so we know whether a prompt or model change made things better before it reaches users.
- Get the right context to the model. Turn user questions into effective queries against our search platform, orchestrate multi-step research loops, and shape the context the model reasons over. When an answer is wrong, work out whether retrieval, the query or the prompt is at fault - and fix the right one.
- Debug production. Instrument flows with tracing (Langfuse), investigate bad answers from real traces, and manage latency, token and cost budgets - including routing across model sizes and families behind an AI gateway.
Requirements
- Experience in building and operating backend services - APIs, async, testing
- Hands-on experience taking LLM features to production and keeping them running - not only prototypes
- Agent / orchestration frameworks - LangGraph ideally
- Practical RAG experience
- Experience debugging LLM systems in production - tracing, evaluation, cost and latency
- Experience running services in the cloud (we're on Azure)
- Strong problem-solving skills, analytical thinking, and technical decision-making
- Fluent in English, proactive communicator, and a collaborative team player
- Open-minded, creative, and motivated to push boundaries in AI and automation
Nice to have:
· Azure OpenAI, AI Search, App Service
- Infrastructure-as-code (Bicep)
- Mentoring or tech-lead experience
Tech stack:
· Python
- FastAPI
- LangGraph
- Azure OpenAI, Google Gemini, Anthropic Claude
- FAISS
- PostgreSQL
- Langfuse
- Azure App Insights
- Bicep
- Docker/Podman
Benefits
- B2B contract
- 100% remote work
- Long-term engagement
- Work on a live LLM product
- Modern AI/LLM technology stack
- Flexible working environment
Originally posted on Himalayas