8 working systems. All local-first. All live.
"Press play looking at someone. Glasses say: Alex Chen. Met 3 times. March meetup. Talked about her startup."
DynamoDB — multi-device sync, works away from home
Transcribe Streaming — 200ms ASR, speaker ID built in
Bedrock — Claude in AWS infra, IAM auth
Lambda — recall API when laptop is off
Lambda + API Gateway → drop-in ingest endpoint
Kinesis → real-time gaze stream, thousands of users
SageMaker → population-level calibration model
Personalize → gaze + emotion → adaptive content
Lambda — sync preset libraries across devices
Personalize — per-user gesture profiles
IVS — stream live gesture performances
Bedrock — cloud AI backend, no local model needed
Lambda — sync session state and bindings
CodeWhisperer — voice/gesture triggered suggestions
Rekognition — cloud fallback for low-confidence face matches
SageMaker — train population-level expression classifiers
Kinesis — stream per-frame emotion events at scale
Transcribe Medical — clinical ASR, medical vocab
HealthLake — structured consultation records
Comprehend Medical — NLP on clinical notes
Browse properties as head-tracked 3D walkthroughs. Place transparent verified bids directly to owners. No agents. No commissions.
S3 + CloudFront — 3D model and splat delivery
Lambda — bid processing, notifications
SageMaker — Gaussian splatting pipeline
These systems are working today — locally, privately, without infrastructure cost. Cloud infrastructure turns each one from a prototype into a product. Below is what becomes possible with the right backing, and why each one makes sense.