AI Backend Engineer
Antrino Labs AB · Stockholm · Tills vidare · Deadline 2026-11-20
Antrino Labs AB is hiring an **AI Backend Engineer** to build the core intelligence layer of its vision intelligence platform. Operating between the model development team and product engineers, the role focuses on developing the backend services that transform customer intent into executing systems. Key responsibilities include creating Python and FastAPI services for detection and measurement, building the compiler and runtime that convert interface graphs into executable model plans, establishing routing and cost-gating logic, and engineering agentic flows that let language models manipulate structured graphs. The engineer also designs APIs, manages PostgreSQL data models with row-level security, integrates proprietary and frontier models, and debugs complex production issues spanning orchestration and model calls. The position involves collaborating with engineering teams in **Stockholm and Bangalore**, with potential business travel required. Applicants must possess **substantial production backend experience**, ideally where machine learning models or LLMs form a core component of the system logic. Candidates need strong proficiency in Python, comfort working with TypeScript or Node.js, and practical experience designing APIs, data boundaries, and PostgreSQL databases. Essential qualifications also include experience utilizing LLM APIs for tool use, structured output, or multi-step agentic pipelines, alongside an understanding of testing non-deterministic systems. Experience with real-time or streaming data is a strong advantage. Fluent **written and spoken English** is required, while proficiency in Swedish is considered a plus.
We are looking for an experienced AI Backend Engineer to build the core intelligence layer of a fast-scaling AI company backed by NVIDIA, Microsoft, and AWS.
About Antrino
Antrino Labs is building the platform that makes every square meter intelligent.
The physical world runs on processes nobody can actually see. Goods move, machines idle, space goes unused, and decisions about all of it are based on samples, guesses, and reports written after the fact. Antrino lets any organization deploy vision intelligence into a physical space and get back what is actually happening there, as structured, real-time information it can act on. Not a research project or a custom integration, but a platform where you describe what matters and deploy it.
We have grown from 6 people to a team of 28 in less than a year, closed our Seed round, announced together with Dagens Industri, and are registered with Inspektionen för strategiska produkter (ISP), the Swedish authority for strategic and dual-use products. Everything we build follows Privacy by Design.
The role
You will build the services that turn a customer's intent into working intelligence.
When a customer builds something in Studio, our composition canvas, that graph has to become a running system: a plan that decides which models to call, in what order, gated by cost and confidence, and that produces events and evidence a customer can trust. You will work on the compiler that turns the graph into an execution plan, the runtime that executes it, and the Python and TypeScript services that detection, classification, and measurement run through.
You will also work on our Agent, which interviews customers, proposes and edits these graphs, and deploys them with confirmation. Getting an LLM to reliably manipulate a structured graph, validate it, and explain its reasoning is a real engineering problem, not a prompting exercise.
The role sits between the team building our first-party detection models and the engineers building the product interface. It is backend systems engineering for a product whose core logic is built out of AI calls.
What you will work on
- Backend services for detection, classification, and measurement, largely in Python with FastAPI
- The compiler that turns a Studio graph into an execution plan, and the runtime that deploys and runs it
- Routing logic that decides which model handles which work, including cost gating where a cheap model decides when an expensive one runs
- The Agent's interview, graph-editing, and deployment flows that turn natural language into a working pipeline
- Integrating our first-party detection models, and frontier model APIs where they do something ours do not yet, behind one clean interface
- Turning raw model output into structured, auditable events and evidence an operator can trust
- APIs and data models for streams, behaviors, runs, and events in PostgreSQL with row-level security
- Tests and evaluation harnesses for logic that includes non-deterministic model calls
- Debugging production issues that span a customer request, our orchestration, and the model call underneath
Our stack
You do not need experience with all of this, but you should recognize most of it and be able to argue about it.
Backend: Python, FastAPI, TypeScript, Node.jsData: PostgreSQL through Supabase, with row-level security,
Streaming: RTSP ingestion, RTSP→HLS conversion, live and archive pipelines
Infrastructure: Azure, GitHub Actions, on-site compute reached over Tailscale
- Who we are looking for Substantial backend experience on production systems, ideally where an LLM or ML model was part of the core logic rather than a bolted-on feature
- Strong Python, and comfort in TypeScript or Node.js
- Real experience designing APIs, data models, and service boundaries, including PostgreSQL
- Experience building with LLM APIs beyond a chat wrapper: tool use, structured output, agentic flows, or a model as one step in a larger pipeline
- An understanding of how to test systems with non-deterministic components
- Experience with real-time or streaming data is a strong advantage
- Excellent written and spoken English. Swedish is a plus, not a requirement
You can reason about a graph or compiler-shaped problem even if you have never built a compiler. You make decisions with incomplete information, take responsibility for the outcome, and say so clearly when something you built was wrong.
Life at Antrino
You will join a team of driven, ambitious people who thrive at a fast pace and enjoy working together. We go to AI events, host our own, and end plenty of weeks with an afterwork. You will work closely with our engineers in Stockholm and our growing team in Bangalore.
What we offer
- Direct access to the founders, the engineering team, and customers
- Real ownership of a core layer of the platform rather than delegated tickets
- Potential participation in our stock option programme after the probationary period
- Hardware of your choice
- AI events, afterworks, team activities, and travel when the business requires it
- Occupational pension, insurance, and friskvårdsbidrag
Process
1. Introductory conversation about your background and the role
2. Technical conversation about systems you have built and why you made the decisions you made. No whiteboard algorithm puzzles
3. Working session with the engineering team on a real problem from our stack
4. Conversation with the founders about options and terms
We review applications continuously. References may be requested at the final stage.
Apply with your CV and, if you have one, a link to something you have built. A repository, a shipped product, or a short note about a system you are proud of tells us more than a cover letter