Latest AI JobsBrowse Jobs
Cognition logo

Research Engineer, Mid-Training

CognitionSan FranciscoResearch & Development
Apply on company site
Salary
Not listed
Job type
Full-time
Posted
6 Aug 2026
Status
Open · checked 7d ago

About this role

Role Overview

The Research Engineer, Mid-Training role at Cognition in San Francisco involves shaping late-stage model training decisions to build advanced software agents. This position suits individuals with deep technical expertise in large language model pipelines who want to work in an applied AI lab setting.

Key Responsibilities

Design and iterate on high-quality data mixtures for late-stage and annealing training runs, developing principled methods for sourcing, filtering, and weighting data.
Drive targeted improvements in coding, mathematics, and long-horizon reasoning through curated data strategies and training interventions.
Develop and evaluate synthetic data pipelines that generate training signal at scale.
Research and optimize multi-stage learning rate schedules, warmup strategies, and compute allocation across training phases.
Research and implement methods for extending effective context length without degrading short-context performance.
Build evaluations that distinguish real capability improvements from benchmark overfitting.
Measure how mid-training interventions scale with compute and data, developing new approaches when existing methods hit ceilings.

Requirements

Deep familiarity with the end-to-end LLM training pipeline, covering pre-training data, optimization, architecture, and the interaction between mid-training and post-training.
Hands-on experience with continual pre-training, annealing, or late-stage data mixing for large models.
Strong intuition for data quality, including dataset curation, large-scale filtering, and the compounding effects of data mix choices.
Experience developing or evaluating synthetic data pipelines for capability improvement.
Strong fundamentals in optimization, statistics, and machine learning theory.
A track record of original contributions through publications, open-source impact, or internal results.
Ability to operate comfortably in an ambiguous, fast-moving environment.

Skills and Tools

Python
PyTorch
Large Language Models (LLMs)
Distributed training frameworks

About the Company

Cognition is an applied AI lab creating end-to-end software agents and the creator of Devin, the first AI software engineer. The team brings together competitive programmers, former founders, and leaders from top AI organizations. The environment emphasizes speed, autonomy, technical depth, and minimal process overhead with large GPU compute allocations.

How to Apply

Interested candidates should apply directly on the official Cognition careers page.

Skills and tech stack

PythonPyTorchLLM

Questions candidates ask

Where is the Research Engineer, Mid-Training position located?

The role is based in San Francisco, California.

What is the primary focus of the mid-training role?

The role focuses on late-stage training decisions, including data mix and quality uplift, annealing schedules, context length extension, capability injection, and synthetic data strategies.

What programming languages and frameworks are required?

Candidates need proficiency in Python and deep learning frameworks such as PyTorch, along with experience debugging distributed training at scale.

What kind of environment does Cognition provide for research?

Cognition operates as a small, highly selective team where research and product move together. Compute is not a constraint, with training jobs routinely running across thousands of GPUs.

How should I submit my application?

Candidates can apply by visiting the official Cognition careers page.

Ready to apply?

The application is on the official Cognition careers page.

Apply on company site