AI 应用全栈工程师

About the role
We are building the internal AI platform that product teams across the company build on. It is the layer between our data, AI and engineering teams and the models they use, whether those come from external providers or run on our own GPUs, and through those teams it reaches thousands of businesses and millions of users.
You will own work end to end: the API, the services behind it, the data model, the interface people operate it through, the tests, and how it behaves in production under real load.
What you will do

• Design and build secure REST APIs and real-time streaming endpoints that other engineering teams depend on daily

• Build backend services that stay correct under concurrency, partial failure, and traffic spikes, and that hold a 99.9% availability target through peak business hours

• Build web interfaces that make a complex system legible, so an operator can understand state and act on it without reading source code

• Model data carefully in PostgreSQL and write queries you can defend on cost as well as correctness

• Build and run a unified API gateway over external model providers and the self-hosted models you deploy and operate on our own GPUs, with low-latency routing, load balancing, and failover, so the teams who consume our APIs never see the differences between backends

• Build multi-tenant boundaries that hold: authentication (OAuth2), role-based access control, quotas, and rate limits that fail closed rather than leak

• Measure usage and cost accurately enough to report and bill from

• Instrument what you ship, and use that instrumentation during incidents to find the real cause rather than a plausible one

• Take part in delivery: code review, CI/CD, progressive rollout, and the debugging that follows a bad release

• Work closely with Product Managers and the internal teams who consume the platform to turn their needs into solutions they actually adopt

• Work with the Technical Program Manager to run the development lifecycle: concept, design, test, release, and support

• Work closely with DevOps to operate and maintain the platform in our infrastructure

• Keep team knowledge written down: technical requirements, API contracts, deployment notes, and post-mortems

• Mentor other engineers and raise the bar on design and code quality through review

Requirements
Engineering

• 4+ years of software engineering experience in a team setting, building and running production web applications

• Strong JavaScript and TypeScript, with production Node.js experience

• Production experience with at least one modern frontend framework. Vue is preferred; React or Next.js also works, and we will expect you to become effective in Vue regardless of which you arrive with.

• Working knowledge of Go language

• PostgreSQL in production: schema design, indexing, transactions, and diagnosing a slow query rather than guessing at it

• REST API design, plus practical experience with streaming responses and long-lived connections

• Testing as part of the change rather than a later cleanup (TDD or close to it), and comfort with code review as a two-way conversation

• Good understanding of microservices design patterns and where they cost more than they return

Production and infrastructure

• Demonstrated ownership of scalability and reliability in high-traffic systems, including API gateways, load balancing, and operating against availability and error-rate targets

• Security fundamentals in day-to-day work: OAuth2 and role-based access control, credential handling, tenant isolation, input validation, and least privilege

• Docker, and container orchestration with Kubernetes and Helm

• CI/CD pipelines and Git-based workflows, including release and rollback

• A public cloud in production. Experience with Alibaba Cloud, AWS, GCP, or Azure.

AI application experience

• You have shipped LLM-backed features to real users, not only prototypes

• Familiarity with multiple model providers and their APIs (OpenAI, Anthropic, and others), including aggregators, and an understanding of where their contracts differ in practice rather than in documentation

• Practical grasp of streaming responses, tool and function calling, embeddings and retrieval (RAG with a vector database), multimodal input, and provider batch and file APIs

• Experience deploying and operating self-hosted models in production (LLMs, embedding, speech-to-text, text-to-speech, or multimodal) with an inference server such as vLLM, SGLang, TGI, or Triton, including the trade-offs between GPU capacity, latency, and throughput

• Experience building agent workflows that automate multi-step processes, and knowing where they need guardrails

• Some way of telling whether model output is actually good, whether that is evaluation sets, human review, or production signals

• Awareness of cost and latency as product constraints, not afterthoughts

• Working proficiency with AI-assisted development tools such as Claude Code or Codex

Collaboration and ways of working

• Experience collaborating directly with product teams and AI engineers on technical development of features and services

• Familiarity with Scrum and Kanban

• Strong written and verbal communication, with a habit of sharing context with teammates and stakeholders

• Ability to build and deploy solutions independently, from problem framing to production

Nice to have

• Experience deploying models across multiple GPUs or multiple nodes, or fine-tuning models for a specific use case

• Experience with microfrontend architectures (e.g. Module Federation, single-spa)

• Experience with Ruby frameworks (e.g. Rails, Sinatra)

• Experience building internal developer platforms or APIs consumed by other engineering teams

Originally posted on Himalayas