AI x TechBio: Featuring the future of drug discovery | MMC

AI x TechBio: Featuring the future of drug discovery

user<br>Advika Jalan<br>, Charlotte Barttelot

icon-yoi 08.06.26

Data-driven health

AI x TechBio series

Part 1

Part 2

Insights Hub

AI x TechBio: Featuring the future of drug discovery

In Part I of our AI x TechBio series we drew on 40+ interviews with senior pharma leaders and startup founders to break down what “good” looks like across (1) proprietary data, (2) algorithms, and (3) lab-in-the-loop infrastructure and agentic AI workflows, illustrated with case studies spanning new architectures like JEPA to curiosity-driven agentic AI systems for novel discovery. We also outlined how startups are validating their AI models early, even before clinical trial data is available.

While Part I covers the technology, Part II turns to the commercial landscape and particularly the startups innovating in it.

We map c.120 AI x TechBio startups and scaleups, highlighting diversity across therapeutic areas, modalities, workflow stages, biological approaches, and business models. There are most certainly overlaps amongst the categories – for instance, Basecamp Research sits in "Nature-derived discovery" but also belongs in "gene therapy." That’s why, in addition to the visual map, we’ve created the tabular version, which highlights the multiple therapeutic areas and modalities that a startup is working with. We also explore how pure platform TechBio companies (without their own drug pipelines) can still build large, valuable businesses.

Note: Certain startups and scaleups may fall into multiple categories and sub-categories; for the sake of simplicity we have assigned each company a single sub-category. We have primarily focused on AI x TechBio startups founded 2016 onwards (the only exception being Formation Bio, which is the exemplar of a particular business model).

The market overview

AI co-scientists

01.

AI co-scientists are emerging as intelligent, agent-based platforms that augment or automate large parts of the scientific process – from hypothesis generation and literature review to experiment design, execution, and analysis. Rather than point solutions, these systems act as integrated “scientific collaborators,” combining reasoning models with tool orchestration, data integration, and workflow automation to run end-to-end research cycles. They enable faster iteration, institutionalise knowledge (so they capture and compound organisational scientific knowledge across the discovery lifecycle), and shift scientists’ focus from manual execution to higher-level thinking and decision-making. Startups in this space include DaltonTx, Kiin Bio, Coincidence Labs, Phylo, Potato, Edison Scientific, GXL, Science Machine, and Causaly.

Bio AI infrastructure

02.

As the number of biological AI models increases (from structure predictors to foundation models for genomics, transcriptomics, and single-cell data) a bottleneck is emerging: most bench scientists cannot use them without ML engineering support, GPU provisioning, and bespoke data pipelines. Tamarind Bio addresses this by hosting 200+ open-source and community models (AlphaFold, RFdiffusion, GROMACS, and others) in a no-code cloud platform. Meanwhile, Helical provides both a unified framework for working with bio foundation models and its own pre-trained models with a Virtual Lab for biologists and a Model Factory for ML engineers to personalise and fine-tune models on proprietary data. Salt AI provides a platform to deploy, build, optimise, and run multi-model systems for drug discovery workflows.

Modern labs

03.

The lab of the future is increasingly autonomous, software-defined, and deeply integrated with AI – where scientists describe goals and intelligent systems design, run, and iterate experiments end-to-end. Platforms like b12 and Briefly translate natural-language instructions into executable protocols across connected lab hardware, while Adaptyv Bio and OnePot AI provide cloud or automated labs that physically synthesise and test proteins or small molecules at speed, feeding high-quality data back into AI models. Instance Bio and Dash Bio represent the digitised backbone of this ecosystem, turning experimental and clinical workflows into fully automated, data-rich pipelines that accelerate bioanalysis and drug development.

Reactwise applies Bayesian optimisation and pre-trained chemistry models to process development, integrating with robotic lab systems to cut up to 95% of experimental work in chemical scale-up. Meanwhile, Differential Bio combines robotic lab automation with mechanistic and data-driven AI models to optimise bioproduction scale-up, tackling manufacturability bottlenecks in fermentation and microbial cell culture. Ganymede Bio provides a cloud PaaS ("Lab-as-Code") that unifies instruments, LIMS, ELNs, and other sources into a single data layer, while UniteLabs standardises the proprietary interfaces between instruments...