Lead optimization in small-molecule research takes months, and most drug candidates never make it to clinical testing, so Merck needed a way to predict molecular behavior earlier and reduce wasted lab work.
Reported impact
Merck reports that AI is accelerating early drug-development work by helping scientists identify disease targets and optimized compounds faster, improve candidate quality, reduce costs, and shorten preclinical phases.
Merck built KERMT (Kinetic GROVER Multi-Task) with NVIDIA to support the lead-optimization stage of small-molecule discovery, where researchers refine early compounds in search of a viable medicine. The core compute element is a deep-learning foundation model trained on more than 11 million molecules and large volumes of chemical data. In the workflow described, molecular data is used to train KERMT, the model learns chemical patterns, and Merck scientists then use its predictions to assess how candidate molecules may behave in the body so liabilities can be detected earlier than with lab work alone. The page does not describe the storage or orchestration stack in detail, but it does say the open-source model is available on NVIDIA accelerated computing and software, including NVIDIA BioNeMo and Clara Open Models, providing a platform path for packaging and serving the model to Merck researchers and the wider scientific community.
Chemical data from more than 11 million molecules is used to train KERMT. The model learns molecular patterns and is then applied by Merck researchers during lead optimization to predict how candidate compounds may behave in the body, helping teams surface issues earlier and focus lab work on the most promising molecules.
"AI is sometimes cutting timelines by 30% or more, improving drug candidate quality and reducing costs."
"Developed in collaboration with Nvidia, KERMT, pronounced 'Kermit' and short for Kinetic GROVER Multi-Task, is a deep-learning computer model trained on more than 11 million molecules."
"It learns from patterns in vast amounts of chemical data with the goal of helping scientists better predict how a molecule will behave in the body, potentially spotting issues much earlier and reducing the need for months of lab work."
"KERMT is available on Nvidia accelerated computing and software, including platforms like Nvidia BioNeMo and Clara Open Models."
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
Related
NVIDIA Sample Sep 22, 2026 DELFI Diagnostics DELFI Diagnostics needed to speed genomic analysis for cancer screening and therapeutic monitoring, but CPU-bound alignment and somatic variant-calling steps created turnaround-time and cost bottlenecks that limited how… 10–20 business days DeepSomatic · MutectCaller · NVIDIA Parabricks NVIDIA Sample Jul 10, 2026 Astellas Pharma Inc. Astellas needed to shorten the slow handoff between experimental scientists and computational chemists in early drug discovery so more researchers could use structure prediction and affinity screening without waiting on… More than 50 regular users and over 2,000 total users within four months NVIDIA BioNeMo · NVIDIA Boltz-2 NIM · NVIDIA DGX H100/H200 NVIDIA Sample Apr 23, 2026 Biofy Hospitals needed pathogen identification and antibiotic-resistance analysis much faster than five-day culture and antibiogram workflows, but CPU-bound genomic processing and limited AI expertise made it hard to turn… 100,000+ synthetic DNA sequences generated by the biology model Abby · H100 Tensor Core GPUs · NVIDIA CUDA-accelerated libraries 