MI / 01
Machine Intelligence
Representation learning, world models, scientific ML and task-specific architectures for problems where generic models are not enough.
MODEL / LEARN / PREDICT
IW / 00 Engineering intelligence
Models for the physical world.
Systems built to be trusted.
Izdeliya Works is a technical works for machine intelligence, physical systems, and secure computing — where research is expected to survive contact with reality.
01Machine
Intelligence
Intelligence
02Physical
Systems
Systems
03Secure
Computing
Computing
ENGINEERED
INTELLIGENCE
INTELLIGENCE
01
Areas of work
One engineering practice.
Three connected domains.
We work at the boundary between learned models and engineered systems. The domain changes; the standard does not.
PS / 02
Physical Systems
Learning from measurements of real systems — including EIS, degradation, materials and environmental state — toward useful digital twins.
MEASURE / STATE / EVOLVE
SC / 03
Secure Computing
Hardware-backed identity, trusted interfaces, secure transports and defensive systems designed around explicit boundaries and minimal attack surface.
IDENTITY / TRUST / BOUNDARY
02
Current work
PRODUCT / 001ACTIVE DEVELOPMENT
Izdeliye-231M
Izdeliye-231M is a secure hardware-authentication bridge that lets a Mac use a physical FIDO security key through a nearby iPhone, with private credentials that never leave the hardware key.
FIDO2 / WebAuthn
Encrypted device link
macOS + iPhone
PIN protected
Discoverable credentials
Multi-assertion
HOSTmacOSCTAP-HID
BLE
BRIDGEiPhoneAuthenticated
NFC
TRUST ROOTFIDO keyPrivate key stays here
VIRTUAL AUTHENTICATORPHYSICAL CREDENTIAL
RESEARCH / PHYSICAL INTELLIGENCEIN DEVELOPMENT
World models for engineered systems.
Our AI/ML direction centers on representations and predictive models grounded in measured physical processes. Initial work focuses on EIS and corrosion: learning useful system state from electrochemical observations, environment and material history, then connecting that state to diagnosis, prediction and digital-twin workflows.
03
Method
Research when necessary.
Engineering always.
Novelty is useful only when it removes a real constraint. We prefer the simplest mechanism that can survive the environment it is built for.
Find the hard constraint.
Identify the security, learning or physical-system limitation that actually controls the outcome.
Model what matters.
Use existing components when they are sufficient. Research new representations, mechanisms or architectures when they are not.
Build the whole system.
Integration, deployment, failure modes, observability and operational constraints are part of the design, not cleanup work.
MEASURE BEFORE CLAIM
MODELS MUST EARN COMPLEXITY
TRUST BOUNDARIES ARE EXPLICIT
BUILD FOR THE REAL ENVIRONMENT
IW / CONTACT Research · products · systems
Build something
that has to work.
For research collaboration, product development, or technical enquiries.
izdeliya.works@kawach.ca ↗