Menu

01 — What I’m exploring

Areas / work and research

What I’m exploring.

Areas of work, research and development — not a list of things I claim to have mastered. The through-line is human identity and autonomy: technology that protects the person holding it.

Read the areas

02 — The areas

Ten areas

Areas of work, research and development.

  • Voice interfaces

    Building and testing conversational systems — and asking what a voiceprint should never be allowed to become.

  • Digital identity and voiceprints

    Ways to prove a property about a person without turning their face or voice into a permanent key.

  • Proof of Human

    Proving aliveness without handing a vendor the one credential that cannot be reissued.

  • Blockchain

    Where an immutable record helps, and where it only makes a bad decision permanent.

  • AI safeguards and security

    Limits that hold when the model is useful, fast and confident — and protection for the system and the person on the other side.

  • AI orchestration

    Making several models behave as one accountable system.

  • Safe data management

    Collect what the task needs, keep it only as long as the task needs, and be able to prove absence.

  • Memory Block

    A persistent memory that belongs to the person, not the platform.

  • Unrestrained model experimentation

    What I call unhinged models: pushing systems past their intended use to find where the safeguards actually are.

  • Frontend architecture and advanced JavaScript

    The layer where all of it either works for a person or does not.

03 — The through-line

Human identity and autonomy

Technology should protect the person holding it.

The areas look separate. Voice, identity, proof, security, memory, orchestration: they are the same question from different sides. Who does the system serve, and what does it take from the person to work?

I am interested in technology that empowers rather than replaces. Systems that prove a property without extracting a person. Memory that belongs to the person, not the platform. Security that holds when the model is confident and wrong.

I do not claim to have solved any of it. I claim to be working on it, in public where I can.

04 — The invitation

If you are building any of this, I want to hear about it.

Voice, identity, proof, security, memory — or something I have not thought of yet.

Write to me