The leap, lived.

What you'll work on

In your first year, paired and supported, you can expect to —

• Write and review Terraform alongside a senior engineer, until infrastructure-as-code becomes a language you think in.
• Help stand up and tune observability — dashboards, alerts, and the hard-won difference between noise and signal.
• Shadow on-call before you ever carry a pager, then carry one when you and your lead agree you're ready.
• Take part in cloud migrations, including the unglamorous, essential work of making them safe.
• Learn one cloud platform deeply — GCP, Azure, or AWS — and enough of a second to translate.

What we look for — the craft

• A genuine foundation in how computers and networks work.
• Hands-on exposure to one cloud platform — a free tier, a certification, an internship, a deployed project.
• Some scripting fluency — Python, Bash, Go, or similar.
• A first encounter with infrastructure-as-code — Terraform, Pulumi, or CloudFormation.
• Something you've built and kept running, that you can walk us through — including the parts that broke.

What we look for — how you work

• AI as a force multiplier, from day one. You already use Claude, Cursor, Copilot, or agents to write faster, debug harder, and learn quicker.
• Comfort being uncertain in the open. We hire for the ability to say "I don't know yet" and then go and find out.
• Care for the people downstream. You can write a clear note and explain a problem to someone who isn't an engineer.
• Mistakes are part of the work — never the measure of you.

What we look for — the craft

• Deep fluency in at least one major cloud platform; working knowledge of at least one other.
• Infrastructure as code (Terraform, Pulumi, or equivalent) — not just to deploy, to reason with.
• Strong opinions on observability — what to instrument, what to alert on, what to silence.
• Familiarity with security and compliance regimes (SOC 2, ISO 27001, RBI guidelines, equivalents).
• A track record of running production systems that someone is paying for.

What we look for — how you work

• AI as force multiplier, not replacement. You use Claude, Cursor, Copilot, agents — to write IaC faster, debug harder, document better. We'll teach you the next layer together.
• Comfort being uncertain in public. We talk about what we don't know in standups; we don't pretend.
• Customer-grade communication. You can write a clear runbook and explain a P0 to a CTO without jargon.

What you'll work on

In your first year, paired and supported, you can expect to —

• Help build retrieval pipelines, and the unglamorous data work that decides whether they are any good.
• Write and maintain evaluations — because a system is only as trustworthy as the eval set standing behind it.
• Wire up agents and copilots against real customer workflows, and watch how they behave outside the demo.
• Help instrument AI systems in production — latency, cost, drift, and the many ways a model quietly fails.
• Learn the modern stack hands-on — model gateways, vector stores, agent frameworks — on work that ships.

What we look for — the craft

• Comfortable in Python — you read other people's code and write your own without copying it whole.
• A real understanding of how modern AI gets used — you've called an LLM API, built retrieval, or fine-tuned a model.
• A first encounter with the modern toolchain — a vector database, an agent framework (LangGraph, the OpenAI or Anthropic SDKs), or a model gateway.
• Some instinct for evaluation — you've felt the discomfort of not knowing whether an output is good, and tried to measure it.
• Something you've built with AI you can demonstrate — including what broke and what you changed.

What we look for — how you work

• AI as a force multiplier, from day one. You use AI to help build AI — code, evals, data labelling.
• Comfort with stochasticity. Models are random, and so is much of this work.
• Comfort being uncertain in the open. We hire for "I don't know yet — let me find out".
• Care for the people downstream, and the understanding that mistakes are part of the work.

What we look for — the craft

• Production ML / AI engineering — model serving, evals, observability, retraining loops.
• Comfort across the modern stack: PyTorch, transformers, vector DBs, agent frameworks (LangGraph, OpenAI Agents, Anthropic SDK), model gateways (Bedrock / Vertex / Azure AI / OpenAI / Anthropic API), embedding pipelines.
• Working knowledge of at least one cloud platform's AI services.
• Strong opinions on evaluations. You believe a system is only as good as the eval set you've published.

What we look for — how you work

• You're already using AI to build AI — multi-agent systems for evals, agentic code review, LLM-augmented data labelling. Show us how.
• Customer empathy. You can sit with a CIO and explain why a particular eval failed and what we'll change.
• Comfort with stochasticity. Models are random. So is the work.

What you'll work on

In your first year, paired and supported, you can expect to —

• Build and maintain data pipelines alongside a senior engineer, and learn why a pipeline is not done when the bytes arrive.
• Help model data — turning raw operational tables into something an analyst, or an AI system, can rely on.
• Write the tests and checks behind data contracts, until "is this number right?" has an answer you trust.
• Take part in reconciliation — making the dashboard agree with the source system, and stay agreed.
• Learn a modern data stack hands-on — warehouses, transformation tools, orchestration — on work that ships.

What we look for — the craft

• Comfortable with SQL — joins, window functions, aggregations.
• Some Python fluency — enough to move and reshape data, not do it by hand.
• A real understanding of what a data pipeline is; you've built one and watched it run more than once.
• A first encounter with the modern data stack — a warehouse, dbt, or an orchestrator (Airflow, Dagster).
• Something you've built that handles real data, that you can walk us through — including the messy parts.

What we look for — how you work

• AI as a force multiplier, from day one. You use Claude, Cursor, Copilot, or agents to write SQL and Python faster.
• A feel for data as a product, not a side-effect — you care whether the numbers are right and documented.
• Comfort being uncertain in the open. We hire for "I don't know yet — let me find out".
• Care for the people downstream, and the understanding that mistakes are part of the work.

What we look for — the craft

• Production fluency in modern data stacks — Snowflake, BigQuery, Databricks, dbt, Airflow, Dagster, Spark, Kafka, or your equivalent of these. Depth in some, fluency in more.
• Cloud foundations — IAM, networking, cost, recovery — on at least one cloud.
• Strong opinions on data contracts, observability, and documentation. You think of data as a product, not a side-effect.
• Willingness to work where the gap is — modeling one quarter, latency the next, cost the quarter after.

What we look for — how you work

• AI as force multiplier. LLM-assisted SQL, agentic ETL, semantic auto-mapping — tools you reach for, then teach the rest of us.
• Comfort being on the customer's side of the table. You'll often work embedded with a client team for weeks.
• The instinct to make data legible to humans downstream — engineers, analysts, executives. All of them.