PostEra – Machine Learning Researcher

Role Overview

In this position, you’ll develop the agentic analysis vertical at PostEra, utilizing agentic programs to automate the event of mechanistic fashions for biochemical and physiological processes, and analyse organic knowledge for goal validation. You will work together intently with chemists and biologists to make use of these fashions to drive drug discovery choices. 

You may also develop machine studying strategies that may quickly adapt to new drug discovery issues from restricted labeled knowledge. A specific focus is molecular and tabular in-context studying and constructing basis fashions from PostEra’s proprietary multimodal knowledge. You will construct fashions that use the context of drug discovery successfully, decide which prior examples and duties are related, quantify when switch is useful or dangerous, and supply dependable predictions beneath distribution shift. 

You will assist drive the complete analysis loop: defining duties, developing datasets and analysis episodes, creating sturdy baselines, coaching and scaling fashions, performing rigorous ablations, and translating profitable strategies into capabilities utilized by PostEra’s scientists. Prior drug discovery expertise just isn’t required, however try to be motivated to study the area and work intently with medicinal chemists, computational chemists, and different scientists.

Key Responsibilities

• Agentic Research for Chemistry and Biology: Develop and benchmark agentic programs to automate the event of quantitative fashions to simulate organic and physiological processes, and the evaluation of organic knowledge. 

• Research Direction and Execution: Independently establish, formulate, and lead analysis tasks involving in-context studying, agentic programs, few-shot adaptation, tabular basis fashions, and molecular machine studying.

• In-Context Learning for Molecules: Design and practice fashions that adapt to new assays, endpoints, targets, or chemical collection utilizing restricted labeled context and heterogeneous historic knowledge.

• Benchmarking: Rigorously evaluate new approaches towards sturdy baselines, curating check instances that deconfounds impression of various sources of bias. 

• Cross-Functional Collaboration: Work with scientists to attach modeling targets and analysis metrics to sensible choices in efficiency modeling, ADME prediction, selectivity, lead optimization, and the design of early medical research.

• Model Training and Scaling: Develop environment friendly coaching and knowledge pipelines and, the place applicable, scale fashions throughout giant collections of molecular and tabular duties.

• Research Engineering: Produce readable, reproducible analysis code; preserve well-tracked experiments; and contribute via code evaluate, documentation, and shared modeling transport systems.

• Scientific Dissemination: Publish ends in main machine studying, medicinal chemistry, or computational biology venues. You can be an envoy of PostEra to the scientific neighborhood. 

Key Requirements

• PhD diploma in machine studying, or STEM analysis involving the event of novel machine studying approaches 

• Track report of high-quality analysis, corresponding to publications and open-source contributions

• Strong analysis or engineering expertise in fashionable machine studying, deep studying, or statistical modeling, backed by understanding of the idea behind machine studying algorithms

• Demonstrated experience in at the very least one related space: agentic workflows for science, machine studying approaches to bioinformatics and medical knowledge modelling, in-context studying, tabular studying, few-shot studying

• Hands-on expertise coaching, debugging, and evaluating ML fashions in Python utilizing frameworks corresponding to PyTorch or JAX

• Ability to independently translate ambiguous scientific or technical issues into well-defined ML tasks, together with datasets, activity definitions, baselines, metrics, and validation schemes

• Ability to design cautious experiments, benchmarks, and ablations that distinguish enhancements from biases, and perceive which features of the mannequin led to the enhancements

• Comfort working in a startup atmosphere the place priorities evolve, knowledge is imperfect, and high-quality judgment issues as a lot as uncooked mannequin complexity

Nice to Haves

Candidates are extra probably to reach this position if in addition they have expertise with a number of of the next:

• Training tabular basis fashions, significantly for sparse, heterogeneous, small-data, or high-missingness settings

• Developing molecular in-context studying programs or adapting general-purpose in-context fashions to molecular or scientific knowledge

• Large mannequin coaching, together with 1B+ parameter fashions, distributed coaching, sharding, knowledge parallelism, mannequin parallelism, and large-scale knowledge pipelines

• The growth of AI “co-scientist” programs for bodily or organic issues 

• Hands-on expertise in modelling organic, biochemical or medical knowledge utilizing machine studying approaches 

• Moving analysis fashions into manufacturing scientific software program or computational workflows

Salary: 200k – 300k

Equity: 0.05 – 0.1%

Visa Sponsorship: Not at the moment

We use recruiting software program to assist the hiring course of, corresponding to accumulating, organizing, and managing purposes. Applications are reviewed by our recruiting staff, and we don’t use synthetic intelligence or automated decision-making instruments to guage candidates or make hiring choices. Human judgment is used all through the hiring course of, and all last choices are made by our hiring staff. If you prefer to extra details about how your private knowledge is processed, or about your rights beneath relevant knowledge safety legal guidelines, please contact us.