À propos de ce poste E11 Bio - Machine Learning Scientist I/II chez Convergentresearch
E11 Bio is on a mission to advance connectomics towards the scale of whole mammalian brains. We are a collaborative and interdisciplinary team of scientists and engineers engaged in a neuroscience moonshot project to develop a radical technology platform for scalable brain mapping. E11 Bio is a non-profit Convergent Research focused research organization located in Alameda, California, pursuing blue-skies neuroscience research yet operating like a nimble, tight-knit start-up. Read our technology roadmap.
We are seeking a creative and mission-driven Machine Learning Scientist to develop scalable automatic proofreading methods leveraging barcode and shape information. In this role, you will own core ML pipelines for automated proofreading of neuron and synaptic reconstructions in optical connectomics datasets, working closely with wet lab scientists, microscopists, and domain experts to ensure biological fidelity and scalability. Your work will directly support E11’s goal of making reliable connectomic data accessible to the broader neuroscience community.
In this role you will be focused on decreasing segmentation and proofreading costs for optical connectomics datasets that we are generating at E11, a critical bottleneck in the connectomics field. You will be responsible for improving initial segmentations using large spatial contexts with rich data features, and ingesting resulting datasets into scalable proofreading software for dissemination to the scientific community.
Responsibilities
Contribute to developing innovative ML approaches for automatic proofreading of segmentations in large PB-scale optical connectomics volumes, leveraging shape and barcode information
Own the deployment of scalable workflows to ingest large connectomic datasets into standard proofreading software (WebKnossos, CAVE)
Collaborate with academic and industry experts on building and extending tools for automated proofreading and interactive visualization
Develop, implement, and evaluate scalable pipelines for synapse detection and partner assignment in large-scale optical connectomics datasets, integrating morphological and protein labeling information
Identify and build ML-based image processing tools that remove key bottlenecks in our data processing and analysis pipeline
Maintain excellent laboratory science practices, communicate results through reports and presentations, and contribute to and assist in preparation of manuscripts, abstracts and other scientific documents.
Required qualifications
Ph.D. in computer science, machine learning, applied math, computational biology or a related field in science and/or engineering or Ph.D.-equivalent publication history
3-5 years of relevant work or research experience, with strong history of independent and autonomous execution of novel ML solutions
Strong background in computer vision and applying ML tools to the analysis of large microscopy data sets
Experience with software engineering best practices (e.g. version control, code reviews, etc)
Excellent communication and collaboration skills, record keeping ability, and attention to detail
Preferred qualifications
Experience with graph, sequence, or point based ML methods (GNNs, Transformers, Pointnets, etc.)
Experience with processing large chunked data (e.g Zarr) in the cloud (e.g. AWS or GCP)
Experience with large scale connectomics data and tools (e.g. Neuroglancer, WebKnossos, CAVE, CATMAID, etc.)
Strong record of publications in peer reviewed scientific journals a plus.
In the first six months, you will:
Become familiar with our existing ML and distributed processing code bases
Own the processing (segmentations) of a new dataset end-to-end and identify areas for improvement (accuracy and efficiency)
Integrate internal graph-based methods for automatic proofreading into standard block-wise workflows for PB-scale optical connectomics volumes
Optimize ingestion of smaller segmentation graphs into standard proofreading software (e.g WebKnossos, CAVE) for improving efficiency of ground truth generation
In 12 months, you will:
Develop novel innovative ML approaches for automatic proofreading of neuron segmentations, leveraging both shape and barcode information
Contribute to algorithmic development across the segmentation pipelines increasing computational throughput
Contribute to achieving cubic-mm scale reconstruction at 100x lower proofreading costs
Scale ingestion of PB-sized optical volumes into proofreading software and pilot proofreading and dissemination of large connectomic subgraphs
You would love this job if…
You are passionate about moonshot neuroscience, and want to broadly accelerate science through long-term commitment to a high-impact project with a tight-knit team
You have a strong desire and intrinsic motivation to work collaboratively and inclusively as part of an interdisciplinary team science effort
You are committed to the ethos of open science, and want to see your work transparently disseminated
You’re mission-driven, and are seeking an environment where you can take long-term risks on moonshot science, shielded from academic or for-profit incentives
You’re seeking a high-leverage, non-traditional opportunity to maximize the impact of your science, establishing relationships with key stakeholders in the neuroscience community
We Offer
An opportunity to help change the world and work with a world-class interdisciplinary team of scientists and engineers on a field-defining moonshot
Excellent medical, dental, and vision insurance through a PPO plan; parental leave.
Generous time off + paid holidays.
401K matching up to 6% of base salary
Ample healthy and energizing food including daily lunches.
Stipend for fitness and wellness activities.
