May 2019 – May 2022 · Portland, OR
ARCQ
CTO / Deep Learning Engineer
Conversational AI agents in the GPT-2/GPT-3 era.
- Role
- CTO / Deep Learning Engineer
- When
- May 2019 – May 2022
- Where
- Portland, OR
- Focus
- AI
01 · The problem
People want to preserve their writings, recordings and memories — and decide, with consent, how those may shape a digital representation of them.
02 · What I built
As CTO, built conversational agents in the GPT-2/3 era with Dialogflow, TensorFlow, Flask and React — an archive of approved materials, recall of relevant memories, and permissions that families and executors control.
03 · Outcome
A consent-first memory archive and conversational model that keeps source material distinct from anything generated.
Highlights
- ✦Consent and permissions at the center
- ✦Memory recall from an authorized archive
- ✦Executor and family access controls
What I did
- ✓Historical and assistive virtual agents with API services and React UIs.
Toolbox
PythonFlaskDialogflowTensorFlowReact
early LLM days