Building the CareKind Investor MVP
Understand, build, test, and explain CareKind’s coming investor-MVP work—from the fixed-camera edge pipeline through evidence-based cloud review, phone and screen-duration policies, SMS delivery, evaluation, and the later ready-to-install system—while making sound product, safety, privacy, and engineering decisions.
6 sections · 28 lessons
Course outline
The Product and the Proof
- Podcast: What the Investor MVP Must Prove
- Podcast: Evidence Before Conclusions
- Workshop: Read the Roadmap as a Dependency Graph
Inside the Edge Pipeline
- Podcast: One Camera Stream, Many Time Scales
- Podcast: Action Areas, Screen Areas, and Coordinate Geometry
- Podcast: Designing a Sensitive Candidate Gate
Physical Events and Cloud Review
- Podcast: Capturing Hand Hits Without Calling Them Abuse
- Podcast: Detecting Falls When Doll Pose Disappears
- Podcast: The Frame the VLM Never Saw
- Workshop: Two Reviews From One Physical Candidate
Phones, Screens, and Time
- Podcast: Choosing the Second Pretrained Edge Model
- Podcast: Prolonged Visible Phone Use
- Podcast: Screen Exposure Without Claiming Attention
- Workshop: Deterministic Duration State Machines
Turning Detection Into a Product
- Podcast: Possible Event—Reviewing
- Podcast: Dana’s Review and the Automatic Demo
- Podcast: Protected Evidence and SMS
- Podcast: From Mac Demo to Ready to Install
Evaluation, Iteration, and Fundraising
- Podcast: Diagnose the Stage That Failed
- Podcast: Labels, Regressions, and Honest Accuracy
- Podcast: Expert Review Is Not Evaluations
- Podcast: Why the Rehearsal Queue Got Noisy
- Podcast: Fair Replay Needs Frozen Identities
- Podcast: B1 Stops Ordinary Activity From Opening Windows
- Podcast: B2 Admission Is Not a Precision Gain
- Podcast: When to Train a Temporal Classifier
- Podcast: Running the Work Through Linear and Orca
- Capstone: Rehearse and Explain the CareKind MVP