Principal ML Scientist – Predictive Toxicology

About Apheris

At Apheris, we power federated data networks in life sciences to enable the development of machine learning models that outperform what any single organisation can build alone. Biopharma can only synthesise and test a finite number of compounds across assays, structural biology workflows, and complex ADME and toxicity endpoints, limiting the performance of models trained on in-house data.

The Apheris product addresses this by hosting networks where biopharma organisations collaboratively train higher-quality models on their combined proprietary datasets without sharing the underlying data. Our federated computing infrastructure, with built-in governance and privacy controls, ensures that data IP and ownership always remain with the data custodians

About the role

• We're looking for an experienced (principal) scientist to own and grow our expansion into small molecule predictive toxicology and quantitative biology.
• This is a hands-on scientific leadership role: you'll set scientific vision, lead customer and consortium conversations, and integrate real scientific workflows into our platform, turning ambitious scientific goals into models that get used in real drug programmes.
• You'll operate with a high degree of autonomy, owning this agenda end-to-end and acting as a scientific counterpart to customers and partners.

About you

What you will do

• Own our expansion into predictive toxicology and quantitative biology. Take the lead as we grow beyond ADME into the science shaping safe, efficacious therapeutics (for example, multi-omics technologies, image-based screening, high-throughput screening and compound-triage cascades).
• Set the scientific strategy. Define how in silico toxicology and quantitative biology workflows come together across our networks, and which endpoints, assays and modelling approaches deliver value in real drug-discovery decisions.
• Decide how best to use relevant data. Bring your understanding of how these techniques and data are generated and embedded in pharmaceutical R&D, and turn it into a clear view of how to extract the most scientific and commercial value from them.
• Span multiple scientific surfaces. Bring depth across the readouts and endpoints that matter for safety and efficacy, from structure-based off-target liability through to pathway-level, mechanistic interpretation and in vivo pharmacokinetics. Integrate these workflows into our platform so customers can run them at scale.
• Build models that matter. Apply federated learning across partner data to deliver models with performance and applicability no single organisation could achieve—and work closely with industrial partners to embed them in real drug-discovery pipelines.
• Lead the scientific conversation with customers and partners, owning scope, evaluation, delivery and adoption in live drug programmes, while shaping the roadmap around genuine scientific and commercial need.

What we expect from you

• Strong deep learning foundations for molecular AI, for example experience with the architectures commonly used for molecular property modelling (e.g. graph neural networks, message-passing and transformer-based models).
• A profile that clearly demonstrates you understand the concerns that drive toxicity assessment in drug discovery — whatever the specific toxicity endpoints you've worked on (for example DILI, cytotoxicity, or micronucleus/genotoxicity imaging readouts).
• Tangible experience building predictive models and driving the adoption of toxicity models in real drug-discovery programmes or industrial R&D pipelines, working closely with teams to get models into pipelines.
• Working knowledge of how RNA-seq, toxicity screens and image-based screens are used in pharma as part of routine HTS and compound triage.
• Scientific leadership excellence: able to set vision, own a scientific agenda, and lead technical and customer conversations independently.
• Comfortable staying hands-on in the modelling while setting scientific direction and mentoring others — this is a scientific leadership role first, with the opportunity to build and lead a team over time.
• PhD or equivalent experience in a relevant field (computational biology, cheminformatics, toxicology, ML, or similar), plus 6+ years applying ML to drug discovery/life science problems.

Nice to have

• Experience with federated learning, privacy-preserving ML, or other multi-party training environments.
• Evidence of prospectively validating predictive toxicity models and using them to influence compound design, prioritisation or progression decisions in live drug-discovery programmes.
• Production-grade model delivery in regulated, enterprise, pharmaceutical, or biotech settings, and/or a publication record in relevant computational biology, toxicology, or ML venues.
• Multi-omics and high-content imaging experience (e.g. cell painting).
• Familiarity with public toxicology and bioactivity data resources (e.g. Tox21, ToxCast, LINCS/L1000) and mechanistic frameworks such as adverse outcome pathways.

What we offer you

• Industry-competitive compensation, including early-stage virtual share options
• Remote-first working – work where you work best
• Wellbeing budget, mental health support, work-from-home budget, co-working stipend, and learning budget
• Generous holiday allowance
• Office Days at our Berlin HQ or a different European location (3x per year)
• A high-calibre, execution-focused team with experience from leading organizations

Logistics

Our mission statement