Srijit Seal

Our Perspective in Nature Reviews Drug Discovery on the current state of AI in the field.

A collaboration between researchers, startups, and venture partners.

Our Nature Communications paper on cell count, cytotoxicity, and benchmark design.

The Polaris collaboration on testing whether models generalize to new chemistry.

Our Cell Systems paper on cytotoxicity and mode-of-action analysis.

Reflections on reviewing datasets and evaluations for NeurIPS.

What I learned from colleagues across discovery science and computational toxicology.

A job asking for more years of PyTorch experience than PyTorch has existed.

Recognition from the SOT Regulatory and Safety Evaluation Specialty Section.

A March 2026 collection of positions in pharma and biotech. Some roles may now be closed.

The published review of applications and case studies from academia and industry.

Our Nature Methods review of the assay, its applications, and the community using it.

Using Codex while thinking through assay biology, validation, and activity cliffs.

Reference managers, DOI checks, and the responsibility to verify citations.

Sharing a free reading link and thanking the review's co-authors.

Celebrating a postponed graduation ceremony and the support behind my research.

An April 2026 collection of positions in pharma and biotech. Some roles may now be closed.

Models for liver toxicity, cardiotoxicity, cell imaging, and pharmacokinetics.

Why imaging models need comparisons against technical controls.

A discussion of preprint policies and concerns about LLM-generated submissions.

Linking image-based profiling to biological measurements and sharing ML-ready datasets.

Questions about pay and working-hour expectations in startup job postings.

Late-night slide preparation and familiar requests from reviewers.

Announcing the June 2023 conference at the University of Cambridge.

A reflection on factual errors in reviews and the use of AI by reviewers.

Statistical precision, model assumptions, and real-world sources of error.

Questions about the quality of future discoveries, as well as their number.

An invitation to discuss foundation models for non-clinical safety.