Software Engineer | AI Training Data & Evals Lab

Atomic - Remote Jobs; Vancouver; Remoto; CAD 111.000 - 159.000; Tempo pieno

Vantaggi offerti da questo lavoro

Fully remote

LATAM focus

US time overlap

Career growth

Impact on AI data

Descrizione del lavoro

Atomic - Remote Jobs is seeking an experienced engineer to lead the design, build, and operation of production evaluation and experiment infrastructure for frontier AI applications. This role blends platform engineering with data pipelines and scalable services, delivering reliable systems used by researchers and operators.

You will own end-to-end systems, ship production improvements within 30–90 days, and collaborate in an async, distributed team across time zones.

Competenze

• Strong coding fundamentals in Node.js and TypeScript.
• Strong coding ability in Python and/or Go.
• Experience building and owning production systems such as APIs, services, and pipelines.
• A solid understanding of distributed systems and engineering trade-offs.
• Comfort with AWS or GCP and modern infrastructure (containers, Kubernetes).
• A track record of shipping and maintaining systems other people rely on, not only prototypes.
• Strong written communication and comfort working async in a distributed team.

Mansioni

• Build and maintain evaluation harnesses that measure model and agent performance on real tasks.
• Improve eval reliability, coverage, and signal quality through better rubrics, task design support, and scoring.
• Ship tools that let researchers and operators run experiments without reinventing the process each time.
• Build APIs and backend services that power human-in-the-loop workflows, task routing, and quality checks.
• Improve the pipelines that turn expert work into structured training and evaluation data.
• Make systems more observable, scalable, and easier to operate through logging, metrics, and debugging.
• Write clear, maintainable code, take part in reviews and design discussions, and document decisions so others can build on them.

Conoscenze

Node.js

TypeScript

Python

Go

APIs

Services

Pipelines

Distributed systems

AWS

GCP

Containers

Kubernetes

System ownership

Communication

Strumenti

Docker

Kubernetes

Descrizione del lavoro

Company Overview

Our client is a fast-growing applied research lab building the data layer for frontier AI. They partner with leading AI labs and enterprises to deliver two things: proprietary, expert-generated datasets, and rigorous evaluation and benchmarking. The goal is for AI systems to get better at real workflows, not just polished demos.

Our client is a fast-growing applied research lab building the data layer for frontier AI. They partner with leading AI labs and enterprises to deliver two things: proprietary, expert-generated datasets, and rigorous evaluation and benchmarking. The goal is for AI systems to get better at real workflows, not just polished demos. Founded in 2025 and backed by a $30M Series A, the company is fully remote. It works on problems such as turning real-world work into clean training signals, building evaluations for software engineering agents and finance workflows, and proving data quality by measuring actual performance lift.

Your Role

This is not a "pick up tickets and wait for specs" role.

• This is a broad builder seat that combines platform engineering with evaluation and experimentation infrastructure. You'll design, build, and run systems in production that researchers and operators depend on every day.
• The company is scaling the pipelines, evaluation harnesses, and training environments that make its work repeatable, and it needs engineers who can own a system and deliver reliably.
• In your first 30 to 90 days, success means shipping at least one meaningful production improvement and becoming the go-to owner of a core system.

You'll:

• Build and maintain evaluation harnesses that measure model and agent performance on real tasks
• Improve eval reliability, coverage, and signal quality through better rubrics, task design support, and scoring
• Ship tools that let researchers and operators run experiments without reinventing the process each time
• Build APIs and backend services that power human-in-the-loop workflows, task routing, and quality checks
• Improve the pipelines that turn expert work into structured training and evaluation data
• Make systems more observable, scalable, and easier to operate through logging, metrics, and debugging
• Write clear, maintainable code, take part in reviews and design discussions, and document decisions so others can build on them

You Bring:

• Strong coding fundamentals in Node.js and TypeScript
• Strong coding ability in Python and/or Go
• Experience building and owning production systems such as APIs, services, and pipelines
• A solid understanding of distributed systems and engineering trade-offs
• Comfort with AWS or GCP and modern infrastructure (containers, Kubernetes)
• A track record of shipping and maintaining systems other people rely on, not only prototypes
• Strong written communication and comfort working async in a distributed team

Bonus Points:

• Experience with evaluation frameworks, experimentation platforms, or ML tooling
• Experience with data pipelines or workflow orchestration
• Experience building internal platforms for operators or research teams
• Experience in early-stage or high-ownership B2B SaaS or platform teams

What's Offered:

• Full-time, fully remote role with a LATAM focus and meaningful overlap with U.S. time zones
• $7,000-10,000 USD/month, based on experience
• Real ownership, with growth into bigger systems, deeper technical leadership, and projects core to how the company scales
• A lean, async-first team that values clear writing, sound judgment, and follow-through
• Research-adjacent engineering at the frontier of AI, alongside practical platform work
• Direct impact on the data and evaluations used by leading AI labs

Interview Process:

• Take-home assignment covering practical engineering and how you communicate decisions
• Application review by the team
• Founding engineer screen, a deep dive on system design, trade-offs, and past ownership
• Work trial on real-world work, focused on execution, quality, and collaboration
• Offer