Hi HN! I'm one of the founders of Portal. We're excited (and nervous) to see what HN thinks about a product we've been working on for about a year. Portal lets you setup a client portal, so that your clients can login on your own website and have a unified place to send messages, share files, make payments, and more.

Before starting Portal we ran an agency. We couldn't find a great solution that streamlines the experience for our clients so we decided to build our own customer portal. After launching, the idea immediately resonated with other agencies so we quickly rebuilt the product so that anyone could quickly set up their own portal. The idea of providing your clients a streamlined portal struck a nerve outside of agencies as well; so we now serve many really interesting companies - chemistry labs, insurance startups, bookkeeping businesses, etc.

Each portal comes with support for messaging, file-sharing, eSignatures, payments, and forms. Most features we built from scratch so that we could prioritize UX and do things like (1) unify notifications, (2) be fully white label, and (3) make it easier to setup automations across features. You can also embed products you already use. For example, many connect Airtable charts to share project progress with clients and others connect Google Data Studio dashboards so that clients have real-time access to analytics. Every portal is customizable so you can set your theme, design your login screen, upload your logos, and connect a custom domain. After you setup a portal, people usually add a "Client portal" button on their marketing website navbar, so that existing clients have a clear sign in path.

If you're curious to learn more, our website is here(https://www.joinportal.com) and our self-serve onboarding flow is here (https://start.joinportal.com/onboarding).

Any feedback or questions welcome! :)

You could but the downsides are: 1/ It doesn't have the ideal sensors (stereo cameras, or other depth sensor), so you'd have limited functionality 2/ Most people wouldn't want to run their iPhone consistently day after day.

I taught myself programming a few years ago and did the ML nanodegree before. My cofounder Neil studied software at Waterloo and was an engineer at Pinterest.

Thanks. I think in a few years we'll see devices that can do everything locally, and cost a reasonable amount. For now, like others have said, some of the models for pose are too complex.

The demo video on our website was done with 2D pose. We’re in the middle of transitioning to 3D.

Our first prototype ran locally end-to-end. To make that work we used a Jetson TX2 ($600 computer), but the performance was abysmal. In a few years, it might be possible to do what we're doing locally and at a reasonable price.

I just think it's unfortunate that every player in the space has to ship their own hardware (and, obviously, force the user into a walled garden so that competitor's products are inconvenient to use.)

Give me (a power user) something I can run in a docker container locally, or on any of my local Macs or PCs.

Or, assume as a constraint that smart homes need to be able to run themselves. I would appreciate if everyone in the space made an effort at detente while the technology matured to a point where a massive privacy and security hole wasn't required in order to have the thing work at all.

We do some preprocessing on-device and run our more complex models in the cloud. It’s similar to Alexa which always listens for the trigger word, but only streams to servers for the few seconds afterwords. While some image data is processed in the cloud, it’s never stored.

It's accurate enough to pick up number of fingers from a fairly long distance (~5m), and it works well in most lighting conditions, including complete darkness.

Actually the lighting was the other question I had in mind. Impressive. When I worked on the Kinect code, lighting was never really an issue we had or wanted to tackle.

I think something that people don't realize until they start using motion sensing technology is that how intuitive it can really be. It might look awkward at first, but once it's seamlessly integrated (ie. stepping into the sensor's view), it's really just like magic. A good enough sensor can pick up and understand very specific motions.

Anyway, good luck guys.

Yes, OpenPose gave us a lot of inspiration. DensePose is similarly impressive and came out recently.

We did the course together but met many years before, in high school in Canada in '07.

We've been using Slite for ~3 weeks. I highly recommend it - much better organization, useful templates, and by far the best UI in the space.