LatchBio

LatchBio is a venture-backed life-sciences software company headquartered in San Francisco, California. It builds cloud infrastructure and tooling that lets computational biologists, bioinformaticians, and drug-discovery teams run, orchestrate, and share biological data-analysis workflows without having to manage their own servers or pipeline glue code. The company positions itself at the intersection of software engineering and the life sciences, describing a workforce of engineers who work directly with scientific end users and ship code that 'will often touch thousands of scientists the same day it's released.' LatchBio's platform is aimed at the modern biology stack: next-generation sequencing analysis, single-cell and spatial transcriptomics, genomics, and the broader 'lab informatics' category that includes data management and analysis orchestration for research and biopharma organizations. The company markets itself as a way for labs to move faster by replacing ad-hoc scripts and fragmented compute environments with a unified, reproducible cloud layer. More recently, LatchBio has expanded into AI evaluation for biology. Its website publishes a series of benchmarks — including scBench, SpatialBench, EpiBench, VariantBench, BioSecBench, and the TxBench family — that measure how well AI agents perform on experimentally grounded biological tasks such as single-cell RNA-seq analysis, spatial biology, epigenomics, genetic variant interpretation, pathogen genomic surveillance, and therapeutic antibody or oligonucleotide discovery. These benchmarks are presented as verifiable evaluations, with associated preprints and reported top scores, reflecting the company's bet that trustworthy AI agents will become central to computational biology. The company is privately held and venture-backed, with backing from prominent life-sciences and technology investors. It competes in a crowded landscape of bioinformatics platforms and cloud lab-data tools, differentiating through developer-friendly infrastructure, close collaboration with scientific users, and an emerging focus on rigorous AI-agent benchmarking for biology.

LatchBio was founded around 2020 in San Francisco by a team of engineers and scientists, including Kenny Workman, with the goal of building better software infrastructure for biology. The company emerged during a period when life-sciences research was generating unprecedented volumes of data — driven by cheaper sequencing, single-cell technologies, and spatial omics — while many labs still relied on fragmented scripts, local servers, and bespoke pipelines. LatchBio's founding thesis was that computational biology needed a modern, cloud-native platform layer analogous to what had transformed software development more broadly. The company built a platform that lets scientists and engineers define, run, and share biological data-analysis workflows in the cloud, handling orchestration, compute, and reproducibility. It targeted bioinformatics teams in academia and industry, as well as drug-discovery organizations that needed to operationalize large-scale genomic and transcriptomic analysis. LatchBio raised venture funding from life-sciences and technology investors and grew a team concentrated in engineering, design, and business development, with a culture emphasizing direct collaboration with scientific end users. As AI models became increasingly capable, LatchBio expanded its mission toward evaluating and improving AI agents for biology. Beginning in late 2025 and continuing through 2026, the company published a series of benchmarks — scBench and SpatialBench (December 2025), EpiBench and SpatialBench-Long (May 2026), scBench-Long and TxBench-PP (June 2026), VariantBench and BioSecBench-Surveillance (July 2026), BioSecBench-Function (August 2026), and TxBench-Oligonucleotide Discovery and TxBench–Antibody Discovery (September 2026). These benchmarks measure agent performance on experimentally grounded tasks across single-cell analysis, spatial biology, epigenomics, variant interpretation, biosecurity, and therapeutic discovery, and are accompanied by preprints and reported scores. This work reflects LatchBio's evolution from a bioinformatics cloud platform into a company also shaping how AI agents are rigorously evaluated in the life sciences.

• 2026

  Released EpiBench and SpatialBench-Long (May 2026).

• 2026

  Released scBench-Long and TxBench-PP (June 2026).

• 2026

  Released VariantBench and BioSecBench-Surveillance (July 2026).

• 2026

  Released BioSecBench-Function (August 2026).

• 2026

  Released TxBench-Oligonucleotide Discovery and TxBench–Antibody Discovery (September 2026).

• 2025

  Released scBench and SpatialBench benchmarks for AI agents in single-cell and spatial biology (December 2025).

• 2020

  LatchBio founded in San Francisco to build cloud software infrastructure for biology.

Developer-friendly cloud infrastructure and AI evaluation for computational biology and drug discovery.

• Latch platform

  Cloud platform for running, orchestrating, and sharing biological data-analysis workflows.

• scBench

  Benchmark evaluating AI agents on single-cell RNA-seq analysis.

• SpatialBench

  Benchmark testing whether AI agents can analyze real-world spatial biology data.

• EpiBench

  Verifiable evaluation of AI agents on epigenomics analysis.

• VariantBench

  Agentic benchmark for genetic variant discovery and interpretation.

• BioSecBench

  Family of benchmarks for AI agents in biosecurity, including Function, Surveillance, and Refusal variants.

• TxBench

  Family of benchmarks for AI agents in therapeutic discovery, including Oligonucleotide Discovery, Antibody Discovery, and PP (preclinical pharmacology).

1. 1LatchBio official websitelatch.bio
2. 2About Us | LatchBiolatch.bio