Alex Iskold

The real tech behind 7 Techstars NYC grads

The real technology often gets lost in the noise of Demo Days.

We talk about funding, growth, markets, but less about the real tech. And it makes sense because technology is an enabler, it is a means to an end, and not a business by itself.

Well, this batch of Techstars NYC is a little different.

This batch of Techstars NYC is not just building incrementally better products. We selected companies that built technology that’s pushing their space forward. It is real tech worth knowing about.

Cartesian invented world’s first 3D printer for electronics, Data Camp embedded R interpreter in the browser, IrisVR built software to help inhabit designs in Virtual Reality, Keymetrics created PM2 – most popular manager for NodeJS, Localize has recreated web application localization, Pilot is delivering internet thats 10x faster than Google Fiber, and Stream has designed a new kind of data store to help build and scale feeds.

We asked each of these companies 3 questions to get more insight about real tech that they built.

Cartesian Co

Ariel Briner, Co-Founder/CEO

Cartesian Co. enables rapid electronics prototyping with our desktop 3D printer for circuit boards.

1. What is your breakthrough innovation?

Our breakthrough innovation is our use of inkjet cartridges and custom inks. What’s special about printing circuits with our technology is that the ink is not only conductive, but also flexible, has excellent adhesion, and can be printed onto a wide variety of surfaces. This means we can print traditional circuit boards, but also more exotic things like paper and fabric.

2. What tradeoffs did you make?

The largest tradeoff we’ve had to face, and are still facing, is striking the balance between having an easy-to-use product, and being on the cutting edge of technology. There is so much innovation that we can make in our space, and things can move so quickly, but we can’t allow that to come at the cost of poor documentation, bad user experience, or lack of communication.

3. What pieces of your technology are really difficult?

The difficult part of our technology is also our breakthrough: the cartridges and ink. We’re pushing the limits of understanding in conductive inks, and we’re using inkjet cartridges in completely new and different ways. This means a lot of experimentation, a lot of failure, and a bit of (very exciting) success.

DataCamp

Jonathan Cornelissen, Co-Founder/CEO

DataCamp is an online data science school, offering hands-on courses using video lessons and interactive coding challenges.

1. What is your breakthrough innovation?

DataCamp offers a web interface for the most popular statistical programming language: R. Our stack includes different open-source frameworks such as AngularJS, NodeJS, Ruby on Rails and of course R itself. Furthermore, we created our own software to analyze the user’s R code and give feedback on how a user can improve his answer for a specific data science challenge.

2. What tradeoffs did you make?

Modularity vs speed of execution. DataCamp was built in a very modular way. It’s relatively easy to add a different data science language such as Python to the DataCamp platform. Building in that modularity came at the cost of increased initial complexity at first. As DataCamp wants to become the go-to platform for learning every data science tool and method, it was important to think this through from the start.

3. What pieces of your technology are really difficult?

Every user on DataCamp gets his/her own sandboxed R session, in which he/she can generate data, create graphs, estimate models, etc. without any emulation or restrictions. The main difficulty is making sure that we can provision DataCamp to hundreds or thousands of concurrent users while keeping it safe and fun for everyone.

IrisVR

Shane Scranton, Co-Founder/CEO

IrisVR builds software to share, edit, and visualize 3D models in virtual reality.

1. What is your breakthrough innovation?

IrisVR is building technology to view and interact with CAD files in virtual reality. 3D CAD software is fundamentally incompatible with VR. Our software pulls out the necessary data from 3D files and pushes it into a real-time game engine that can render the scene properly in a virtual environment. This real-time engine allows us to implement collaboration, interactivity, and realistic visualization in VR.

2. What tradeoffs did you make?

Early in development, we needed to choose between PC-based VR and mobile-based VR. We decided that PC-based VR, like the Oculus Rift, was going to offer the highest fidelity experience for our customers, and we hedged our bets on that platform first. Running VR from a workstation computer also allows our customers to load larger files than could be supported on mobile devices. Since the announcement of HTC Vive, we’re confident that PC-based VR will be the most comfortable experience for quite some time. IrisVR will roll out mobile support in 2016.

3. What pieces of your technology are really difficult?

3D data processing and compression is incredibly difficult. Our customers often have files that range from 5 million to 20 million polygons. To take that much data and render it in a real-time engine is a monumental task. Current generation hardware has a difficult time displaying images at a low latency, high frame rate, and HD resolution – elements required to create comfortable VR experiences and safeguard against motion sickness.

Keymetrics

Alexandre Strzelewicz, Co-Founder/CEO

Keymetrics is a real-time platform for managing and monitoring Node.JS applications. Keymetrics’s open source solution has been downloaded more than 600,000 times.

1. What is your breakthrough innovation?

Real-time management and monitoring of NodeJS applications at high scale. With Keymetrics you can interact with a large number of servers seamlessly and monitor their behavior in real time and over time.

To achieve this challenge we used NodeJS, a technology that is perfectly fitted for networked software. We built it with strong engineering principles including micro service architecture, test-oriented development and nodes-based infrastructure. The visual dashboard has been built with AngularJS to impact changes in real time.


Keymetrics Dashboard monitoring servers around the world.

2. What tradeoffs did you make?

We had great ideas during the program, like a “Sensor Shop.” It’s a place where you can buy sensors to monitor other parts of your system. However, we decided to put it on hold to focus on building more value for our NodeJS users.

3. What pieces of your technology are really difficult?

Our open-source application manager is the product of two years of development in close collaboration with the community. Managing contributions from the community was not easy. Technically, scaling NodeJS applications across all available CPUs was very challenging to get it right.

We experienced high load when we first launched the Keymetrics dashboard. We had to change the database twice and we finally distributed calculation between Keymetrics and our manager.

Localize

Brandon Paton, Founder/CEO

Localize – localization as a service. Localize helps translate your website with one line of code.

1. What is your breakthrough innovation?

Localize is an example of a service that would not have been possible to build 2-3 years ago because browsers were not powerful or advanced enough. Major performance improvements in Chrome / Firefox / IE is what made Localize possible. Localization used to be a massive server-side project for engineers, but Localize removes the need for that by delivering the same end result, through our client-side javascript library.

2. What tradeoffs did you make?

We err on the side of building fewer features, rather than more. This has allowed us to deliver an extremely high quality experience for the features we support, rather than a smorgasbord of features that are only half-good. As we grow, our feature set will expand, but only once we’re able to deliver those features at a quality level that we’re happy with and are able to support 100%. Sometimes this means turning away customers who have custom requirements, but by doing so we’ve gained even more in other areas of the company.

3. What pieces of your technology are really difficult?

The Localize Javascript library was difficult to get right. It needs to work with any website, on broken HTML, and in old browsers. It was very easy to build an MVP, but there are an infinite number of edge cases to handle. Localize has to work well when running in an unpredictable environment. This made it a lot harder than building an application that you can easily test — you can’t just compile the code and see if there are errors. It required a lot of testing on hundreds of websites, weeks of debugging… a lot of patience.

Pilot

Joseph Fasone, Founder/CEO

Pilot is a new kind of internet provider for businesses.

1. What is your breakthrough innovation?

There are a few things we’re particularly proud of. First, we’re able to provide internet access that’s up to 10x faster than Google Fiber using just one strand of fiber. That aside, we’re building software defined networking (SDN) tools to optimize the reliability and quality of the content that our customers want to access, and we provide our customers with a cloud-enabled management platform. Before Pilot, our customers were paying Meraki thousands of dollars a year just to get basic management and monitoring functionality in the cloud. We can provide this functionality with no additional on-site hardware, and at no charge. In the future, this opens up the possibility for a marketplace of virtualized network services.

2. What tradeoffs did you make?

Even in 2015, there are no great alternatives to a carrier-grade router. Early on, we were hoping to reduce costs by completely virtualizing our routing platform and running it atop traditional servers. While it was a great starting point for us, we quickly ran into scalability issues surrounding physical port density, and low packet-per-second throughput. Despite the high upfront cost, we ended up switching to Juniper MX series routers for packet forwarding. We still virtualize the routing intelligence today.

3. What pieces of your technology are really difficult?

Creating a secure API for managing and manipulating hardware across multiple manufacturers is really tough. Internet routing has very little margin for error. When automating network functions, most development time is spent writing unit tests and exception handlers. To fully provision one new customer’s service, our platform runs over 40 tests to ensure that there are no conflicts on the network.

We provide our customers with usage analytics. Bandwidth sampling and analysis is incredibly hard at scale. Usage reports over a 30 day period can involve sampling billions of packets. To stay ahead of this, we’re constantly running a set of MapReduce queries on a local Hadoop cluster. The outputs are de-identified and securely stored while the original samples are discarded. Once requested, the data is served back to the user using Chart.js on the frontend.

Stream

Tommaso Barbugli, Co-Founder/CTO

Stream is a new kind of data store designed for building, scaling and personalizing feeds. Stream’s Open source solution has been downloaded over 250,000 times.

1. What is your breakthrough innovation?

Stream is a data store designed from the ground up for building, scaling and personalizing feeds. Current in-house solutions use existing software that does not scale, is expensive, and has high maintenance and development costs. Stream uses a highly specialized storage layer that does not have all the shortcomings and limitations posed by traditional data storage software.

2. What tradeoffs did you make?

When building our first version of the product, we had to choose what we were going to build: a richer set of features or a specialized data layer for feeds. We decided that building a rock solid storage layer was the best way for us to achieve our long-term vision. Whenever a tradeoff has to be made, we always look at how that impacts our long term vision and our customers.

3. What pieces of your technology are really difficult?

Delivering high performance services at scale is very hard. Stream is used by hundreds of apps with millions of users. Every part of our infrastructure is designed from the ground up for high availability and low latency.