SaaS to On Premise Migration – Biotech Company
Learn how in less than six weeks, we eliminated their $200,000/year SaaS bill by delivering a custom on-premise ML backend.
[ CHALLENGE ]
OmniAb was facing decreasing growth due to:
Reliance on overpriced SaaS, which drained R&D funding.
Competitors using equivalent services, resulting in commoditization of OmniAb's products.
Limited scalability due linear pricing structure.
The company needed a rapid way to decrease cost without sacrificing security, efficiency, and accuracy.
[ SOLUTION ]
OmniAb partnered with our team to migrate their entire machine learning backend away from LandingLens to a secure, on-premise system in less than 6-weeks. Here's how we did it:
Automated Training, Evaluation, and Iteration
By automating each model's experimentation pipeline, we were able to run dozens of experiments and optimize maximize each model's accuracy - surpassing even LandingLens' multi-million dollar models
Built-in Data Indexing to Reduce Infrastructure Costs
In order to prevent increased computational costs from large microscopy imaging data, we integrated the same indexing methods used in training GPT-5. This allowed our models to rapidly identify and process images that researchers are interested in with having to rely on six-figure GPU infrastructure.Automatic Escalation
In the rare event the AI system was unable to locate a viable citation to an answer, the AI would not hallucinate and would instead automatically escalate the question directly to the founder for an immediate email response.
[ RESULTS ]
-85%
Decrease in cloud-based ML spending
-22%
Decrease in budgeted on-premise ML infrastructure
+6%
Increase in model accuracy compared to SaaS implementation
