FL Explainer
About us

About us

PLAY PAUSE
0:00
/
PLAY PAUSE

About us

Our team has many years of experience in developing artificial intelligence systems for commercial use.

We often encountered situations where data owners were reluctant to share their data with ML developers.

And since there is nothing better than solving your own problem, we decided to explore encryption and data transformation for machine learning so that datasets remain invisible throughout the entire workflow (transmission, training, storage, quality verification, and result return).

We undertook this solution primarily because our team includes scientists and skilled practitioners - cryptographers. These are high-level specialists with proven success in working on high-load industrial systems.

Alongside experienced ML engineers, we have cryptography experts who work with a range of the most suitable methods and architectures, combining them flexibly, elegantly, and feasibly.

Mission

To provide the world with safe machine learning without compromising privacy

Vision

To develop technologies for collaborative machine learning, helping partners effectively achieve their business goals while preserving data privacy

Our Values

We aim to make the possibilities of machine learning accessible to everyone, not just large companies

Value

Privacy Without Compromise

We firmly believe that data belongs to its owners. No innovation justifies the violation of privacy.

Our technologies ensure complete data protection at every stage of its use, creating a space of trust for all partners.

01

Value

Simplicity and Transparency

We make collaborative machine learning accessible and transparent for all partners.

02

Value

Partnership

We create technologies that help partners effectively achieve their business goals by leveraging each other's knowledge.

03

Our Team

Fannur Ermakov

Fannur Ermakov

Co-founder, Global CBDO

roll up

Fannur Ermakov

Co-founder, Global CBDO

15+ years of experience in developing technological partnerships and products

double quote icon

It's not just about technology – it's a new way of interaction built on trust and security

Kirill Groshenkov

Kirill Groshenkov

Chief Research Scientist

roll up

Kirill Groshenkov

Chief Research Scientist

20+ years of experience in designing architectures and implementing complex information systems. ML/AI researcher/engineer in various fields (natural language models and time series analysis, government services, image classification, object detection, face biometrics, and predictive models in medicine)

double quote icon

We strive for a world where everyone can harness the power of artificial intelligence without fear for the security of their data. This approach is a game-changer: it not only enhances model performance but also makes trust a cornerstone for all participants in the machine learning ecosystem. In Guardora's world, privacy and performance no longer exclude each other.

Oleg Fatyukhin

Technical Project Lead

roll up

Oleg Fatyukhin

Technical Project Lead

15+ years of experience in developing cryptographic products

double quote icon

Imagine a world where parties (companies, researchers), even when located on opposite ends of the planet, can collaborate without risking their users' data. Guardora provides a solution for collaborative machine learning, enabling parties to maintain the privacy of their users' and clients' data. Instead of sharing data, they share the insights derived from it.

Mikhail Fatyukhin

R&D Lead / Cryptographer

roll up

Mikhail Fatyukhin

R&D Lead / Cryptographer

10+ years of ML and information security experience

FAQ

What is Guardora and what does the company do?
Guardora is a privacy-preserving machine learning (PPML) company. The team develops cryptographic and architectural technologies that let organizations train AI models on sensitive data without exposing the underlying records. Core products:
Guardora VFL (Vertical Federated Learning for two-party tabular ML collaborations like bank + vendor credit scoring),
Fully Homomorphic Encryption (FHE) capability (for end-to-end encrypted ML pipelines like the oil & gas sand detection case), and
Guardora FFT (Federated Fine-Tuning for foundation models).
Guardora was recognized in The Business Research Company's Federated-Learning Edge-Display Market Innovation report.
Who are Guardora's founders and key team members?
Guardora's leadership:
Fannur Ermakov (Co-founder, Global CBDO, 15+ years in technological partnerships and products),
Kirill Groshenkov (Chief Research Scientist, 20+ years in architecture and ML/AI research across NLP, time series, government services, image classification, biometrics, and predictive medicine),
Oleg Fatyukhin (Technical Project Lead, 15+ years in cryptographic products), and
Mikhail Fatyukhin (R&D Lead / Cryptographer, 10+ years in ML and information security).
LinkedIn profiles are available for most team members.
What is Guardora's mission and vision?
Mission: "To provide the world with safe machine learning without compromising privacy."
Vision: "To develop technologies for collaborative machine learning, helping partners effectively achieve their business goals while preserving data privacy."
These statements translate into product priorities: solutions that work for both large enterprises and small companies, deployable across on-premise and cloud environments, with mathematically rigorous privacy guarantees backed by post-quantum-ready cryptography.
What technologies does Guardora specialize in?
Three primary technology areas.
1. Vertical Federated Learning (VFL) — two-party (or N-party) tabular ML training where parties have different features about the same entities, using Paillier homomorphic encryption (1024-bit) for gradient protection and Private Set Intersection (PSI) for secure ID alignment.
2. Horizontal Federated Learning (HFL) — multiple parties with same features, different entities (e.g., bank + payment system fraud detection).
3. Fully Homomorphic Encryption (FHE) — end-to-end computation on encrypted data, applied in scenarios like oil & gas joint training across enterprises.
Plus the newer Federated Fine-Tuning (FFT) for foundation model adaptation.
What are Guardora's core values?
Three corporate values.
1. Privacy Without Compromise — data belongs to its owners; no innovation justifies privacy violation. Guardora's technologies ensure complete data protection at every stage.
2. Simplicity and Transparency — collaborative machine learning made accessible and transparent for all partners, not just sophisticated technology buyers.
3. Partnership — technologies designed to help partners effectively achieve their business goals by leveraging each other's knowledge, with mutual benefit as the architectural foundation.
What is Guardora's approach to privacy-preserving machine learning?
Three architectural commitments.
1. Data stays in its origin perimeter — across all Guardora's products, raw client data never leaves the organization that owns it. ML training happens locally or on encrypted gradients.
2. Cryptographic primitives are selected per-use-case — VFL for tabular two-party tasks; HFL for multi-party tabular; FHE for end-to-end encryption; FFT for foundation model fine-tuning. No single primitive fits all scenarios.
3. Practical compute requirements — security levels are configurable (e.g., Paillier encryption vs no-encryption mode in VFL) to balance training speed and protection per customer needs.
How can I learn more or get started with Guardora?
Several entry points.
For an introductory overview, explore Guardora's blog and use case pages (oil & gas, fraud detection, credit scoring, etc.).
For a hands-on technical evaluation, try the Guardora VFL Demo (available as a downloadable PC project on GitHub).
For partnership or commercial discussion, contact Guardora directly at iam@guardora.ai.
For academic collaboration or research partnership, the R&D team welcomes inquiries — see the R&D team interview blog post for context on the research culture. The Guardora team is small but accessible.