- Case Studies
- /
- AI Model Packaging & IP Protection for Drug Discovery Company
AI Model Packaging & IP Protection for Drug Discovery Company
Secure, cross-platform delivery of proprietary ML models with full IP protection
Faster packaging process
IP protection of proprietary AI models
Our team developed a toolkit to combine trained AI/ML models, inference code, and runtime into an independent obfuscated executable.
Check TestimonialBiotech
Industry
Poland
Location
Secure ML model delivery, IP protection, ML obfuscation, Dockerized AI
Services
$50,000 to $199,999
Budget
The most impressive and unique about this company was their exceptional commitment to customer satisfaction.
Serge H
Head of Partnerships of Software Development CompanyA biotech enterprise needed a secure, standardized way to distribute machine learning models to their clients — without exposing proprietary source code, model weights, or internal processing logic.
See what we can do for youSolution
We rebuilt the entire packaging pipeline into a modular, secure, and portable system for black-box AI model delivery. The solution allows the client to distribute encrypted, obfuscated, and fully containerized ML models with flexible execution options.
Let’s talk about what’s possible
To deliver a encrypted, obfuscated, and fully containerized ML models, Blackthorn AI applied:
Project duration
01-03 Weeks
Migrated legacy 2-file system into modular OOP structure. Added linters, formatters (pylint, mypy, bandit, black) for long-term maintainability. Structured codebase for future extensibility.
04-06 Weeks
Built CLI around Pipeline–Action design pattern. Added YAML configuration support. Developed internal state-sharing mechanism for pipeline steps.
07-09 Weeks
Integrated PyArmor and Nuitka into pipeline with configurable toggles. Designed file encryption as reusable pipeline step with backward compatibility.
09-10 Weeks
Enabled Docker-in-Docker support for cross-platform compilation (macOS, Linux, Windows). Addressed previous limitations of Linux-only builds.
11-12 Weeks
Built unit, integration, and end-to-end test suites. Achieved 90%+ coverage. Delivered stable MVP for internal rollout and team onboarding.
Team Size
Delivering Impact
100%
IP protection of proprietary AI modelsAll model packaging workflows were redesigned to ensure zero exposure of company background IP, including pre-processing scripts, proprietary architectures, and internal logic.
3×
Faster model packaging processTime required to prepare and secure each model for client delivery was reduced by over 65%, dropping from multiple days to just a few hours per model.
>90%
Test coverage across execution FlowIntegration, unit, and e2e tests were implemented across all key steps (obfuscation, encryption, Dockerization). This allowed safe scaling and updates without regressions.
