ML Research Engineer, Evaluation

PassFort; Greater London; On-site; GBP 65,000 - 100,000; Full time

Job summary

Novogaia is seeking a machine learning research engineer to design and run rigorous benchmarks for molecular AI systems. You will own evaluation pipelines, construct adversarial tests, and translate results into clear research priorities for the team.

You will collaborate with AI engineers, computational biologists, and chemists to ensure benchmarks reflect real discovery challenges and provide defensible conclusions about model utility.

Qualifications

• Experience building or evaluating ML benchmarks.
• Excellent technical communication to researchers and non-technical stakeholders.
• Ability to analyze model behavior critically.
• Strong Python proficiency and containerized pipelines.
• Familiarity with cheminformatics formats (SMILES, InChIKey).
• Interest in computational mass spectrometry.

Responsibilities

• Develop a deep understanding of Novogaia's models, data, and evaluation needs.
• Design benchmark tasks reflecting real discovery problems across multiple modalities.
• Build datasets and controls ensuring benchmark trustworthiness: hard negatives and leakage-safe splits.
• Communicate evaluation results clearly to the modeling team and document findings.
• Lead evaluation work, define what good looks like, and translate results into priorities.

Skills

Benchmark design

Python programming

Model evaluation

Scientific communication

Mass spectrometry familiarity

Cheminformatics basics

Tools

Docker

Git

MLflow

Job description

Overview

Novogaia is an applied AI drug discovery company. We build machine learning systems that decode the chemistry of natural organisms, starting with fungi, to find the next generation of medicines.

We are a small team of AI engineers, computational biologists and chemists building foundation models for molecular structure prediction from mass spectrometry data. We are seeking a machine learning research engineer who excels at designing rigorous tests for molecular AI systems, and who can turn \"does this model actually work\" into a concrete, defensible answer. You will own the design of our internal benchmarks and evaluation pipelines; build the adversarial checks that catch shortcut learning and leakage. You will work closely with our modeling team to translate evaluation results into research priorities.

The Role

• Develop a deep understanding of Novogaia's models, data, and evaluation needs




• Design benchmark tasks that reflect real discovery problems: de novo molecular structure generation, molecular and spectral retrieval, mass spectrum simulation, molecular formula prediction, and compound property/class prediction




• Build the datasets and controls that make a benchmark trustworthy: hard negatives, leakage-safe splits, and null baselines that catch a model exploiting shortcuts instead of genuine signal




• Communicate evaluation results as clear findings for the modeling team, and as documentation and data cards that others can trust and reproduce




• Work across the team to scope and lead evaluation work, including:

  
  
  
  

  • Defining what \"good\" looks like for a given model or task, and choosing the right test for it

  
  

  • Analyzing model behavior and interpreting results for researchers and non-technical stakeholders alike

  
  

  • Working with engineers to turn one-off analyses into repeatable, reproducible evaluation pipelines

  
  
  
  



• Translate lessons from evaluation work into research priorities, data requirements, and R&D direction

In your first year, you'll take the lead on how Novogaia measures model quality, from benchmark design through adversarial testing and reporting. The tests you build will decide how much weight anyone can put on our models' outputs.

What We Require

• Research or applied experience in machine learning, with direct experience building or rigorously evaluating ML benchmarks




• Exceptional technical communication skills, including the ability to explain evaluation findings clearly to both researchers and non-technical stakeholders




• Ability to analyze model behavior and interpret computational results critically




• Strong proficiency in Python, and comfort with reproducible, containerized pipelines




• Familiarity with cheminformatics representations (SMILES, InChIKey, molecular fingerprints), or willingness to pick these up quickly




• Familiarity with computational mass spectrometry or eagerness to learn

What We Value

• Ability to dive deep into a result until you know whether it's real or an artifact




• Strong scientific judgment and a willingness to question the benchmark's own assumptions, not just the model's




• Motivation to build infrastructure other people can confidently rely and build upon




• Comfortable being the person who tells the team a result doesn't hold up




• Curiosity, low ego, and a willingness to learn the chemistry side quickly, even if it's outside your original training