AI and predictive systems

AI Engineer

Design, train and deploy the behavioural models and LLM-driven journey steps that predict churn, purchase and uninstall and propose the next action, always behind a human approval.

Remote (India)Full-time
Apply for this roleAll open roles

About the role

Every engagement platform ships an assistant now. The winnable ground is the trust boundary: an assistant that drafts a segment, simulates the journey against real history, shows exactly who is affected and waits for approval. You will build the models and the agent loop behind that boundary.

The work spans propensity scoring on the event store, best-time and best-channel learning from each customer's own history, anomaly detection that explains the dimension that moved, and the LLM step that writes, classifies or routes inside a journey. Everything you ship runs under a person's permissions and lands in the audit log.

Responsibilities

  • Design and deploy low-latency inference pipelines that score intent, churn and abandonment signals as events arrive.
  • Build and fine-tune the LLM steps that draft personalised, context-aware messages for WhatsApp, SMS, email and push, with fallbacks and a QA pass before every send.
  • Improve precision, recall and attribution quality across streaming customer journey events, and make every prediction explain itself.
  • Work with the backend engineers to benchmark model latency and cost, and keep inference inside the budget a journey can afford.

Requirements

  • Three or more years deploying production machine learning, including LLM orchestration and agentic workflows.
  • Fluent Python with PyTorch or TensorFlow, and hands-on experience with embeddings and vector stores for ranking or recommendation.
  • Comfort with edge and serverless inference runtimes such as ONNX Runtime or WebAssembly.
  • A habit of measuring: attribution windows, control groups and the difference between activity and lift.

Nice to have

  • Experience with recommendation systems over a product catalogue.
  • Familiarity with Postgres query planning and SQL-defined features.
  • Prior work on human-in-the-loop or approval-gated automation.

What we offer

The basics, done properly.

The same for every role. Written down so it can be held to.

  • Remote-first. The team works from wherever it works best, and most communication is written and asynchronous.
  • A learning budget for books, courses and conferences, spent on whatever makes you better at the job.
  • The laptop and peripherals you choose, replaced when they slow you down.
  • Health cover for you and your dependants.
  • An annual offsite where the whole team meets in person.
  • Flexible hours built around the few meetings that need everyone present.

Hiring process

Four steps, no puzzles, and you are paid for the work.

The whole sequence usually runs inside three weeks. Every stage is with an engineer.

  1. 01

    Application

    Apply from the role page with a link to something you have built. We read every application and reply to each one.

  2. 02

    Technical conversation

    Forty-five minutes with an engineer on the team about systems you have shipped, what broke, and what you would do differently.

  3. 03

    Paid take-home or pairing session

    Your choice: a scoped take-home paid at a fixed rate, or a two-hour pairing session on real code from the platform.

  4. 04

    Offer

    A written offer with compensation and start date, and a call with the engineers you would work with before you decide.

Apply

Apply for AI Engineer.

A name, an address and a link to something you have built is enough. The note is optional.

Your details are used for this application only. See the privacy policy.

Other open roles

Also hiring.

Not the right role yet?

Send a note anyway.

We reply to every application, and we read the ones that do not match a listed role with the same care.

Email careersSee the platform
  • No credit card for the Developer plan
  • Bring your own providers
  • Export everything, any time