# LiteFold

> LiteFold is an applied AI4Science lab and AI drug discovery platform. Its autonomous research agent, Rosalind, works across therapeutic R&D (literature research, protein and molecule design, docking, molecular dynamics, ADMET and toxicity prediction, CMC and process development) on private, customer-controlled infrastructure. LiteFold also publishes open research: the LiteMol-1 molecular foundation model, the AminoWeb protein dataset and the BenchPLM protein language model benchmark.

Key facts:

- Company: LiteFold, an applied AI4Science lab based in Bengaluru, India, working with enterprise life-sciences R&D teams.
- Website: https://www.lite.bio · Platform: https://prod.litefold.ai · Docs: https://docs.litefold.ai · Contact: contact@litefold.ai
- Flagship product: Rosalind, an autonomous AI co-scientist for drug discovery. It reads private context (ELNs, assay databases, sequences, structures, papers, patents), writes and runs code in isolated sandboxes, and calls LiteFold scientific compute without proprietary data leaving the customer's environment.
- Scientific compute: biomolecule and binder design (peptides, antibodies, nanobodies, small molecules, PROTACs), protein structure prediction, protein–ligand docking and virtual screening, pocket and hotspot detection, molecular dynamics, binding affinity, ADMET, toxicity, inverse folding and model fine-tuning on customer data.
- Benchmarks: Rosalind scores 90.24% on BixBench (Future House's benchmark of 200+ real bioinformatics tasks), ahead of K-Dense (90.0%), Biomni Lab (88.7%), Edison (78.0%) and Claude Code with Opus 4.6 (65.3%). LiteFold's MD engine runs the STMV benchmark (1.07M atoms) at 136 ns/day on one RTX 5090, compared with 62 ns/day on an H100. MitoTox predicts mitochondrial toxicity with 82.6% accuracy.
- Security: SOC 2 compliant, workloads run in private enclaves, self-hosted or private-cloud deployment is available, and customers own all IP, models and designs. Customer data is never used to train shared models. Trust center: https://trust.litefold.ai
- Enterprise offerings: a Platform License (1–3 years, self-hosted), a Computation and AI Service (LiteFold designs 10–20 binder candidates with a full workup report) and an End-to-End Service (adds wet-lab SPR validation). Every engagement includes forward-deployed engineers and scientists.
- Backed and supported by: NVIDIA Inception and Mercatus Center (George Mason University).
- Team: Anindya Sannigrahi (CEO & Co-Founder), Cory Kornowicz (CSO & Co-Founder), Siddhant Prateek Mahanayak (Founding Infrastructure Engineer), Nabajit Borah and Aditi Sinha (Founding Computational Biologists), Pannalal Aich (Director of Business Development).

## Product

- [Home](https://www.lite.bio/): Overview of LiteFold and Rosalind: private context, agentic work, scientific compute and specialized intelligence
- [Enterprise](https://www.lite.bio/enterprise): Rosalind for enterprise R&D, how it compares with physics suites and data platforms, security model and engagement options
- [Use cases](https://www.lite.bio/use-cases): Worked examples covering rare-disease variant interpretation, hypothesis generation, lead optimization, docking campaigns, MD stability analysis, biosimilarity, CMC, formulation and preclinical-to-IND work
- [Benchmarks](https://www.lite.bio/benchmarks): Rosalind on BixBench, LiteMol-1 molecular design, MitoTox toxicity prediction and GPU molecular dynamics throughput
- [Documentation](https://docs.litefold.ai): Platform documentation
- [Hybrid Scientific Intelligence Runtime (HSIR)](https://www.lite.bio/blogs/hybrid-scientific-intelligence-runtime): One runtime that combines GPU/CPU workflows, an agent harness, secure sandboxes, an inference engine and a policy layer on customer-owned infrastructure, benchmarked against Modal, Cloud Run, Cloudflare, Daytona, E2B, Vercel and Runloop

## Research

- [LiteMol-1](https://www.lite.bio/research/litemol1): A masked diffusion language model for agent-driven molecular design. One set of weights generates small molecules, linear and cyclic peptides, non-canonical amino acid peptides, depsipeptides, macrocycles and PROTACs, with Monte Carlo Tree Search for multi-objective Pareto optimization
- [AminoWeb](https://www.lite.bio/research/aminoweb): An open-source, ~7.56 TB Parquet collection of 29 cleaned, ML-ready protein datasets spanning sequences, structures, function, evolution and assays
- [BenchPLM](https://www.lite.bio/research/benchplm): A frozen-embedding benchmark of protein language models (ESM-C, DPLM, ESM-2, ProGen). It finds that bigger protein models don't reliably win
- [Hugging Face](https://huggingface.co/litefold): Open datasets and model weights
- [GitHub](https://github.com/litefold): Open-source code

## Case studies

- [Molecule × LiteFold](https://www.lite.bio/case-studies/molecule): How Molecule used LiteFold multi-agent orchestration and Rosalind to design non-canonical peptide binders for two undrugged class A GPCR targets

## Blog

- [Improving Binding Precision of Therapeutic Antibodies with Rosalind](https://www.lite.bio/blogs/improving-binding-precision-of-therapeutic-antibodies-with-rosalind-by-litefold)
- [Rational Design of a Covalent EGFR T790M Inhibitor Using LiteFold](https://www.lite.bio/blogs/rational-design-of-a-covalent-egfr-t790m-inhibitor-using-litefold)
- [Environmental Modulation and Ligand-Induced Stabilization of the β2-Adrenergic Receptor](https://www.lite.bio/blogs/environmental-modulation-and-ligand-induced-stabilization-of-v2-adrenergic-receptor)
- [Ensemble Docking vs Static Docking: When Does Protein Flexibility Matter?](https://www.lite.bio/blogs/ensemble-docking-vs-static-docking-when-does-protein-flexibility-matter)
- [The Generative Geometric Turn in AI Drug Discovery (diffusion models and BoltzGen)](https://www.lite.bio/blogs/the-stochastic-architect-diffusion-models-in-structure-based-drug-design-and-the-boltzgen-paradigm)
- [Molecular Docking vs. QSAR for ADMET Decisions](https://www.lite.bio/blogs/molecular-docking-vs-qsar)
- [Structural Plasticity in the Mutome: Binding Pocket Alteration](https://www.lite.bio/blogs/structural-plasticity-in-the-mutome-mechanisms-of-binding-pocket-alteration-and-therapeutic-intervention)
- [Small Molecule vs Peptide Competition for the Same Pocket](https://www.lite.bio/blogs/small-molecule-vs-peptide-competition-for-the-same-pocket)
- [In-Silico Toxicology Pipelines for Early Drug Candidates](https://www.lite.bio/blogs/tox)
- [Molecular Simulations to Predict Binding Affinity](https://www.lite.bio/blogs/mds)
- [Molecular Docking in Drug Discovery](https://www.lite.bio/blogs/molecular-docking-in-drug-discovery)

## Company

- [About](https://www.lite.bio/about): Mission, story, team and backers
- [Contact](https://www.lite.bio/contact): Talk to the founders or request an enterprise demo
- [Trust center](https://trust.litefold.ai): Security and compliance documentation
- [Status](https://status.litefold.ai): Platform status

## Optional

- [Full text for LLMs](https://www.lite.bio/llms-full.txt): All research, case-study and blog content in one plain-text file
- [Sitemap](https://www.lite.bio/sitemap.xml)
- [More blog posts](https://www.lite.bio/blogs): Intrinsic water in binding, physics meets AI, generative scaffold design, LiteFold DeNovo, bulk structure prediction and AlphaFold primers
- [Releases](https://www.lite.bio/releases)
- [Events](https://www.lite.bio/events)