AI Engineer

Most education today delivers the same lesson to every student, regardless of what they already know or where they are stuck. Companion Labs exists because that is a solvable problem: AI can now teach every student through a genuinely interactive, one-to-one experience, adapting in real time to what each student understands.

As the AI Engineer at Companion Labs, you will own the entire AI layer of Bloom, the company's one-to-one voice-and-whiteboard maths tutor.

You will work across the agent harness, voice pipeline, evaluation systems, personalization, and the infrastructure that makes the experience fast and reliable. The role asks for strong applied AI engineering judgment, production-level coding ability, and the range to move from architecture decisions to shipped code in the same week.

You will work directly with the two founders and a small, close-knit engineering team.

The AI agent harness. Design and build the orchestration layer that drives the tutor's reasoning, pacing, and pedagogical decisions.

The evaluation system. Define how the tutor is measured: accuracy, safety, engagement, and teaching quality across thousands of interactions.

The voice pipeline. Select, integrate, and tune the STT and TTS stack for low-latency, natural conversation with students.

Whiteboard intelligence. Build the system that decides what the tutor draws, writes, and adapts on the shared visual canvas in real time.

Personalization and memory. Own the layer that tracks each student's progress and shapes future lessons around what they know and where they are stuck.

The AI architecture. Work with the founders to define what the system should look like over the next two to three months, then build it.

Built production AI agents end to end. You have shipped an agent system into real user hands, not just prototyped one in a notebook.

Written the harness, not just called the API. You have designed orchestration logic, managed state, and handled failure modes in a live system.

Built and run evals on LLM outputs. You know how to define what good looks like and instrument a system to measure it continuously.

Operated at zero-to-one pace. You have shipped meaningful scope in weeks, made architecture calls with incomplete information, and moved to the next thing.

Comfortable with production-level code. Your background is applied AI engineering with a strong backend foundation, not research or data science.

Curious about how AI systems evolve. You track what is being built in the agent and voice space and form your own views on what the right approach is.

Define the problem before picking the tool. Every build starts with a clear statement of what we are solving and why the current approach is not enough.

Ship in weeks, not quarters. The team moves in short cycles, shows what it has built, and decides the next move from there.

Own the outcome, not just the task. If something is broken or slow, the person closest to it fixes it without waiting to be asked.

Work in the room together. The team is in the HSR office six days a week; the best thinking happens in person, in real time.

Bring the evidence. Decisions are argued from data, user feedback, or a quick experiment, not from instinct alone.

Pull each other's leg. The team is tight, direct, and genuinely enjoys working together; camaraderie is part of how the work gets done.

You work directly with both founders and own the AI layer of the product end to end. The role gives you the technical and strategic foundation of a company built to teach hundreds of millions of students.