AI Inference That
Never Leaks
Your Data
NOMYO wraps every prompt and response in AES-256-GCM + RSA-4096 encryption. Your sensitive data — patient records, financial models, defense briefs — is processed in the cloud as ciphertext. Only you hold the keys.
from nomyo import SecureChatCompletion
client = SecureChatCompletion(
api_key="your-api-key"
)
response = client.create(
model="Qwen/Qwen3.5-35B-A3B",
messages=[{
"role": "user",
"content": patient_record
}],
security_tier="maximum"
)
# Decrypted locally — never stored on server
print(response["choices"][0]["msg"]["content"])Built for regulated industries
Why Enterprises Choose NOMYO
Traditional cloud AI requires you to send plaintext data to a third party. NOMYO eliminates that risk entirely.
End-to-End Encryption
AES-256-GCM
Every prompt is encrypted client-side with AES-256-GCM before it leaves your server. On High and Maximum tiers, plaintext exists only inside the attested enclave — it is never accessible to NOMYO or its host systems.
Per-Request Key Rotation
Fresh key per request
A fresh AES key protects every request and is never reused. An exposed key affects only that single inference — no other traffic can be decrypted with it.
Secure Memory
mlock · zeroing
Our client SDKs can lock request buffers (mlock / VirtualLock) and zero them after use — reporting per request whether locking and zeroing actually succeeded. Server-side plaintext handling is covered by the Data Handling Commitments.
OpenAI-Compatible API
Drop-in replacement
Drop-in replacement for OpenAI's ChatCompletion. Same interface, same tools, zero rewrites. Just swap the base URL and add encryption.
3 Security Tiers
standard · high · maximum
Choose Standard for general use, High for sensitive business data, or Maximum for the most sensitive workloads. Regulated and special-category data (e.g. health records) requires a signed data-handling addendum.
15+ Enterprise Models
0.6B → 48B MoE
From 0.6B lightweight models to 48B Mixture-of-Experts. Includes medical-specialized models (medgemma-27b) and multilingual support (EuroLLM).
Standards-Based Encryption.
Zero Configuration.
NOMYO uses a hybrid encryption scheme that combines the speed of AES-256-GCM for payload encryption with RSA-OAEP (4096-bit) for secure key exchange. The result is a system built on standardized, well-analyzed primitives. All Encrypted Inference runs exclusively on NOMYO-hosted models in our German data center — content is never accessible from outside the facility.
-
Per-request key rotation
Every inference gets a unique AES-256 key. Keys are generated via
secrets.token_bytesand zeroed after use. -
RSA-OAEP key exchange
4096-bit RSA keys establish the secure channel. Server public key fingerprint verification prevents MITM attacks.
-
Memory protection
On High and Maximum tiers, plaintext payloads are protected from swap to disk and memory dumps, and crypto material is zeroed after use. Client-side equivalents with per-request verifiable status are available in the SDKs.
-
Hardware-rooted Attestation
On High and Maximum tiers, inference runs inside an Intel SGX enclave. Your client can verify SGX attestation — opt-in per request — before any plaintext is sent, confirming the exact enclave binary it is talking to. The Maximum tier additionally returns a TPM 2.0 quote covering firmware, Secure Boot and the application binary.
-
Compliance Support
The core property — plaintext never leaves your control — maps directly to central technical safeguards such as GDPR Art. 32 and HIPAA §164.312. These are support claims, not certifications — the binding terms are our Data Handling Commitments and DPA (see the legal section of nomyo.ai).
15+ Models. Zero Setup.
From lightweight edge models to powerful reasoning systems. All models are E2E encrypted by default.
Highest quality general model · 35B total / 3B active params
Google's medical-domain instruction-tuned model · 27B
Largest capacity model · 48B total / 3B active params
ServiceNow reasoning model with chain-of-thought
Strong European language support · 9B parameters
Lightweight, ultra-fast inference · 0.6B parameters
Simple, Transparent Pricing
Start with a single API key. Scale across your organization. No hidden fees, no usage surprises.
For production workloads with sensitive data
- 1 API key
- All models included
- All security tiers*
- 2 req/s (4 burst)
- Professional human support
Maximum security for regulated industries
- Unlimited API keys
- All models + Maximum security tier
- Custom rate limits
- Dedicated support + SLA
- Custom deployments & compliance
* Regulated or special-category data (e.g. health records) requires a signed data-handling addendum — a HIPAA Business Associate Agreement is not currently offered. See our Data Handling Commitments.
Stop Sending Plaintext to the Cloud
If your data is sensitive enough that you can't send it to a third-party AI, you need NOMYO. Get started in under 5 minutes with our OpenAI-compatible Python client.
pip install nomyo