Suraksh AI co-designs AI and encryption, enabling enterprises to deploy fully encrypted models that deliver insights without ever exposing data or IP.
Suraksh AI
Encrypted Intelligence for the Civilization
Summary
Suraksh AI auto-co-designs AI and encryption to deliver insights without exposing data or IP.
As encrypted AI demand surges, enterprises lack usable tools for secure deployment.
We offer:
- AutoFHE Compiler: Push-button encrypted AI compiler
- AutoFHE SDK: Deployment-ready SDK for encrypted inference
- Encrypted Models: Biometrics, LLMs, and genomics
Backed by peer-reviewed research (CVPR, USENIX), Suraksh delivers 10-100x speedups over legacy encrypted AI stacks.
The Problem
Encrypted AI is still out of reach. It is held back by slow tools, vendor lock-in, and custom hardware delays.
Manually designing performant encrypted models is slow, error-prone, and talent-intensive.
Encrypted AI needs solutions today—not years from now on costly, custom silicon.
Enterprises are boxed into rigid hardware and encryption stacks, with no way out.
Billions in secure AI value are stranded because no one fixed the pipeline.
Our Solution
Push-Button Encrypted AI: Faster, Flexible, and Scalable by Design
Accelerate Inference by 10-100x
- Orders-of-magnitude speedup over manual designs
- Latency reduced from 9s to 0.4s (Face model)
- Optimized for CPU inference
Stay Flexible Across Hardware & Frameworks
- No hardware, encryption scheme, or vendor lock-in
- Runs across TFHE, CKKS, BFV, and MPC
- Portable across cloud and on-prem
Scale Secure AI with Push-Button Automation
- One-click encrypted model deployment
- Optimize latency, accuracy, memory tradeoff
- Zero manual tuning required
Our Product
Suraksh AI is the first full-stack platform that co-designs AI models and encryption schemes for truly secure, production-ready encrypted inference.
Push-button compiler that co-optimizes encrypted model architecture and crypto parameters for speed, size, and accuracy.
Deployment-ready SDK that integrates with your existing stack to run fully encrypted inference on ASICs (CPU, GPU, etc.).
Encryption-friendly models for face, fingerprint, genomics, and LLM will be licensed for private inference.
Case Study 1
CryptoFace: World's First End-to-End Encrypted Face Recognition - 7x Faster with Neural Co-Design.
Peer-reviewed and accepted at CVPR 2025—one of the top global AI conferences.
Case Study 2
AutoFHE: Automated Design Beats Manual Experts - Outperforms CryptoLab and Zama
Peer-reviewed and accepted at USENIX Security 2024—one of the top global cybersecurity and systems conferences.
Team
Worked together for 2 years. Have a proven track record in tech startups and academia.
Dr. Vishnu Boddeti
CTO & Co-Founder
- PhD - Carnegie Mellon Univ.
- 20+ yrs. experience in AI
- Expert in Trustworthy ML, AutoML and Biometrics
Dr. Amina Bassit
Founding Engineer
- PhD from Univ. of Twente
- 10+ yrs. experience in cryptography
- Expert in FHE and biometrics
Suraksh AI co-designs AI and encryption, enabling enterprises to deploy fully encrypted models that deliver insights without ever exposing data or IP.
Suraksh AI
Encrypted Intelligence for the Civilization