Modular Workflows for Protein Folding, Co-Folding & Design on AWS

THE PROBLEM: Biotech and pharma researchers face critical challenges in operationalizing protein structure prediction: Managing multiple model versions across research teams Building and maintaining cloud infrastructure (AWS, containerization, orchestration) takes weeks No unified interface for non-technical scientists to submit predictions Custom fine-tuned models remain trapped in research silos Scaling predictions without re-architecting infrastructure Limited cost visibility and optimization Challenges with version control and reproducibility

OUR SOLUTION: END-TO-END BIOMOLECULAR PREDICTION PLATFORM Loka provides an end-to-end platform combining:

STATE-OF-THE-ART MODEL LIBRARY The platform ships pre-integrated with leading open-source and state-of-the-art models for folding, co-folding, and generative protein design, with support for multiple MSA tools. Current integrations include (but are not limited to): Boltz-2 (protein folding, co-folding, binding affinity) Chai-1 (protein co-folding, antibody-antigen interactions) OpenFold & OpenFold3 (open-source AF2 reproduction + multi-modal) RosettaFold (next-generation folding with design capabilities) Protenix (generative + co-folding models) AlphaFold2 (baseline monomer/multimer prediction) Custom Models (bring your own)

DATA PRIVACY & IP All predictions run within the customer's own AWS account. Sequences, structures, and results never leave the customer's infrastructure: no data sharing with third parties. All outputs generated on the platform are owned entirely by the customer. Loka does not retain, access, or reuse any sequence data, prediction results, or fine-tuned model weights.