Scientist I - ML/AI Foundational Models for Synthetic Enhancer Design

alleninstitute; Seattle (WA); On-site; USD 120,000 - 160,000; Full time

Job summary

The Allen Institute seeks a Scientist I to build the computational backbone of a lab-in-the-loop platform that designs synthetic enhancers for brain cell-type targeting. You will develop scalable genomic sequence models, train on cloud GPUs, and prepare data and benchmarks for community use.

You will contribute to open-source projects for training, evaluating, and running inference on sequence-to-function models, enabling reproducible research and collaborative tool-building across regulatory

Qualifications

• Ph.D. in CS, computational biology, bioinformatics, applied math, engineering, or related field, or equivalent combination of degree and experience.
• Strong software engineering practice in Python including version control, testing, code review, dependency management, and reproducible environments.
• Experience training deep learning models in PyTorch, JAX, or TensorFlow on multi-GPU or distributed infrastructure.
• Experience building data pipelines for large scientific datasets that do not fit in memory.

Responsibilities

• Build and maintain an open-source codebase for training, fine-tuning, and serving genomic sequence models, and package models and benchmarks for use outside the Institute
• Train large-scale whole-brain models efficiently in multi-GPU cloud environments
• Turn cross-species multi-omic data into reproducible, versioned training sets
• Implement active-learning loops that feed in vivo screening results back into the next round of models
• Publish and present findings in peer-reviewed journals and at scientific conferences

Skills

Python

Version control

Testing

Code review

reproducible environments

PyTorch

JAX

TensorFlow

Multi-GPU training

Data pipelines

Education

Ph.D. in computer science / computational biology / bioinformatics / applied mathematics / engineering

Equivalent degree + experience

Tools

Git

Job description

Scientist I – ML/AI foundational models for synthetic enhancer design

The Allen Institute accelerates science for a healthier world through large-scale research designed to answer some of the most complex questions in biology. Our multi-disciplinary teams generate foundational knowledge, tools, and data to understand how our brain, cells, and immune system work. We share our work openly so others can build on it, move faster, and ask bigger questions. We drive discovery forward and create new possibilities for improving human health.

Brain Health is a new global collaborative research initiative designed to accelerate understanding of human brain diseases through large-scale human tissue analysis, open science, AI-enabled disease modeling, and translational platform technologies. Building on foundational advances in human brain cell atlases, quantitative neuropathology, single-cell and spatial biology, multimodal molecular profiling, and AI-enabled analysis, Brain Health is developing a scalable framework for understanding disease progression directly in the human brain and identifying new opportunities for therapeutic development.

We seek a Scientist I to build the computational backbone of a lab-in-the-loop platform that designs and validates synthetic enhancers for precision cell-type targeting in the brain. In this cycle, model-designed sequences are tested experimentally, and the results retrain the next round of models. You will contribute to an open-source codebase for training, fine-tuning, evaluating, and running inference on genomic sequence-to-function models. You will scale training across our cloud GPU infrastructure, turn large-scale cross-species whole-brain multi-omic data into reproducible training sets, and build on the thousands of enhancer-AAV vectors already screened and released publicly through the Allen Institute Genetic Tools Atlas to inform the design of novel synthetic enhancers. The ideal candidate is a strong scientific software engineer and ML practitioner excited to share their models, sequences, and code as community tools for exploring regulatory genomics and designing enhancers in any tissue.

At the Allen Institute, we believe that science is for everyone – and should be open to everyone. We are dedicated to combating biases and reducing barriers to STEM careers more broadly.

We also believe that science is better when it includes different perspectives and voices. We strive to make the Allen Institute a place where everyone feels like they belong and are empowered to do their best work in a supportive environment.

We are an equal-opportunity employer and strongly encourage people from all backgrounds to apply for our open positions.

Essential Functions

• Build and maintain an open-source codebase for training, fine-tuning, and serving genomic sequence models, and package models and benchmarks for use outside the Institute
• Train large-scale whole-brain models efficiently in multi-GPU cloud environments
• Turn cross-species multi-omic data into reproducible, versioned training sets
• Implement active-learning loops that feed in vivo screening results back into the next round of models
• Publish and present findings in peer-reviewed journals and at scientific conferences

Note: Reasonable accommodations may be made to enable individuals with disabilities to perform the essential functions. This description reflects management’s assignment of essential functions; it does not proscribe or restrict the tasks that may be assigned.

Required Education and Experience

• Ph.D. in computer science, computational biology, bioinformatics, applied mathematics, engineering, or a related field; or an equivalent combination of degree and experience
• Strong software engineering practice in Python, including version control, testing, code review, dependency management, and reproducible environments
• Experience training deep learning models in PyTorch, JAX, or TensorFlow on multi-GPU or distributed infrastructure
• Experience building data pipelines for large scientific datasets that do not fit in memory<? Actually need closing slider? let me correct: