Coursera is a MOOC pioneer, founded by two leading Machine Learning (ML) professors form Stanford. Yet we don’t see ML used much on Coursera to improve your education. How can it be deployed to radically improve online learning experience? Furthermore, how can ML be used to improve on physical learning?

Here’s an idea I had for a long time. Automatically matching students and teachers online based on performance testing.

Here’s a personal story. I used to have a very tough, very smart Chinese boss. He criticized me a lot. If I were an American, I’d probably just quit. However, I understood that he cares about me, and I wanted to perform what was asked and it was fine. The key is to understand how a teacher relates to a student in a given culture. I’ve actually watched a lot of Chinese movies during that period, and living in the oldest Chinatown in the US (Oakland) helps… But say you’re a student brought up in a liberal tradition, pampered and encouraged all the way. Would you perform better or worse with such a teacher?

The short answer is: we don’t know. A mediocrity carried along by the societal approval can end up finishing an English major at an Ivy league school and inheriting dad’s business and do fine. An immigrant may blossom. Or it may be completely different: the heir would open up to the challenge, and an immigrant would be crushed. One of the first movies I’ve seen fresh off the boat, arriving at Dartmouth in 1993, was The Dead Poets Society. It resonated strongly. Teaching style matters a lot during the formative years, it can set your life’s course. For some, it is a matter of life and death. It is important.

How can we calibrate a student’s response to a teacher? Clearly it is mutually beneficial to both the student and the teacher to find the counterpart working best for them. The success criterion is quite clear: educational progress. So we may formulate it as an ML job!

Here’s how it can go. Say our MOOC site has a 1,000 Spanish 101 courses. We’ll show you 10 most different segments, taught my most different professors in terms of the features we know so far. We’ll give you a test after each segment and measure your performance. One teacher will command you, the other will be nice to you. One will strengthen your resolve, another encourage you gently. Clearly personalities will differ in more ways than what we quantified so far; we’ll learn better features about students and teachers all the time. The good thing is that the profile will apply across courses.

Scrolling through hundreds and thousands of personal styles and optimizing test performance is not something you can do in the physical world. When done properly, this approach can change the way we teach online. I call it Education Engineering.

Today Versal, the edtech company I co-founded, announced its interactive product tutorials. It means the idea I proposed last year is now implemented.

The concept is very simple: most product manuals are static series of pictures. Product videos can only go so far. The really good idea is simulation. Any consumer product with moving parts, knobs, etc., can be simulated as an animated, interactive object in the browser or mobile app. E.g., you will be able to play with a child carseat and make sure you understand how e.g. latch system works.

My other key idea is that interactive objects can be used in two modes:

* instruction
* assessment

Hence, to complete product mastery, we can ask the customer to do certain important things, e.g. remove a safety cover before use, level the car seat, etc.

Another improvement is to add QR codes to all products linked to their interactive tutorials.

A further extension is “real world education”. Imagine that you come to a gym, and don’t know how to operate a machine. Given you are a big, bold guy, it’s embarrassing to ask for instructions. You whip out your smartphone, scan the QR code on the machine for Open Tutorials, and enjoy the simulation in your app!

We’ll probably need to come up with fully open-source standard for the tutorials and simulations, so that manufacturers and consumers are never locked in and the ecosystem can benefit from synergies between both.

A startup called SpotCheck is asking users worried about moles to take a picture and send it to dermatologists for screening.  I think an extension can be putting an iPhone next to your chest to record the audio of the heart and analyze it.

I’m converging on a good working model of the mind economy – a quantitative approach to human behavior which I started with my Ph.D. thesis, the first large-scale practical system modeling human behavior on the Twitter graph. Companies like Quid promised to unravel “startup DNA” after I’ve published the foundations of mind economy, but great intelligence is still locked inside VCs, founders, and generally in the startup ecosystem. The most data-hungry, data-driven crowd does not yet have the tools to fully enable it to discover the “unicorns” of the future, or simply better tools to enable better VC work.

I propose to model people as bags of ideas, companies as bags of people and transitive bags of ideas, and establish an economy which will resolve the value of people, companies and ideas simultaneously, based on the actual exits, or failures, by those companies.

There is a pretty clear plan for the first series of data mining tools, built on Scala platform, ready to power a VC dashboard. A single meaningful investment following from it would justify the very first backing of this project. Multiple further products logically follow from the mind economy agenda.

A whole additional layer of human capital products are naturally complementing the mind economy dashboard. If you’re interested in backing them and being the first users, please contact me at alexy@scalable.pro. If you are a business co-founder or a data scientist interested in quantifying the startup ecosystem, drop me a line – we may get together for coffee in SOMA!

On July 9th, we’ve launched Versal, the edtech platform which is interactive, beautiful, extensible, and functional. It is functional throughout – it runs on Scala.

It has been a thoughtful, serious process of building out the company, it components, systems, and groups. For me, it was transformative, leveraging all of the skills acquired in computer and startup worlds. I’ve been the first coding co-founder to start, and selected Scala as the backend strategy. We’ve built an amazing group of Scala engineers creating beautiful, minimalist code. Our Scala architecture was presented at the ScalaDays conference and attracted a lot of interest already – it is event-sourced and supports the “don’t get married to tech” approach.

I’ve also assembled a fantastic advisory board, spanning the best engineers in industry and academia. We’re engineering education, after all. We’re discovering what it should be.

We are a truly global company. I was lucky to invite my schoolmate Sergei Winitzki to lead our mobile development. He used to be a professor of Physics in Munich, and now is a mobile architect there working for Versal remotely. I’ve also found a fantastic team from Siberia which is enabling us in many ways. We have UI visionaries in Netherlands (where they are from) and Paris (when they are not back home in SF).

Since then, I’ve also had a chance to build the devops group and test several technologies for infrastructure as code, selecting Salt for amazing scalability, responsiveness, and growth capacity. We’re hosting the first Salt hackathon in San Francisco this Saturday. With Salt, I’ve delivered a live migration of the whole stack from VA to OR.

This Summer, I’ve started our internship program, and our awesome interns are paving the way for Versal R&D. It’s amazing to see what we’ve built in just a year. I’ve mentioned just some things where I focused, but the artistic and interactive result is the result of our product, design, and front-end working together with the back-end components. Great things happen when you work with the right technology and the people who self-select with it.

Once the volume of user data increases, we will learn how education really happens, and will help everyone to get the most of it. It will be a beautiful thing.

On April 9, I had an opportunity to teach two classes at Wharton. Prof. Shawndra Hill, who was a member of my CIS PhD committee at Penn, is teaching a course on data mining, and she invited me to present the insights from the startup world to undergrads and MBAs, focusing on data mining in real world as well as operations and everything else which makes a company a reality.

Having bootstrapped operations and backend, as well as staffing and setting strategy for several groups in several companies, it was a great way to summarize some findings so far which could make business majors and MBAs more successful entrepreneurs in the startup world. Wharton has always been a technology-focused business school, but until a few years ago it mostly served large corporations, investment banking, and the like. Despite their business spirit, MBAs from top schools tend to be rather conservative, as our study at TopProspect had shown (based on studying LinkedIn histories of MBAs from the top schools).

In Shawndra’s class, a majority of MBAs had a startup idea or actual startups in various stages of formation! My last slide listed my LinkedIn profile, and by the end of Q&A, I had a dozen requests to connect. MBAs were much more pressing in their questions, but both sections provided many insights.

My two key points were as follows. Startups, SOMA-style, are formed by core founders who want to master an aspect of the world and make it better through their understanding encoded in technology. Learning is at the heart of this endeavor. It is a shared experience of the whole team which plunges into a new area, absorbs everything there is to know about the domain and the technology facilitating it, and discharges a system into the world, like a neuron, which starts propagating new connections and adds value by connecting users with the improved view of the world, even if in the small.

The second, inherently related, and key point, is that the developers building that worldview are key members of the team. A typical fallacy of stereotypical MBAs or “business cofounders” is that they are “idea people,” or “product people,” while the engineers are conveniences for such ideas, or even commodity tools to prototype ideas, get funding, and be replaced by better commodity, “production quality” engineers. I know startups where such a dichotomy had lead to a massive exodus of engineers, after which “idea people” were left with a “careers” page hiring essentially an entire company. (By the way, such startups are excellent sources of talent, so an experience working with them should not be discounted, but may come rather handy for better startups.)

After living in the SOMA startup bubble for a few years, it was rather surprising to me that Wharton folks find many of these beliefs, widely held in the Bay Area, quite new. Being the organizer of the largest Scala meetup in San Francisco and connecting with a community of founders, founding and lead engineers, visionary product managers, designers, and other key characters of the Bay Area, taught me that immersion in the community is the only way to master its ethos and is a prerequisite to building a successful startup. There is a reason why startups thrive in the Bay Area, and whole companies formed in Israel, Russia, or even Malaysia are transplanted en masse here if they find their lucky strides. In the end, this is a people’s business. You learn by experience and by connecting to like-minded folks. The meetups are gamechangers here – there are more that four thousand meetups in the Bay Area, and the majority seems to be about technology (and singles…).

Wharton has a great presence in San Francisco – I pass by its building on the Embarcadero every day, and see its banners advertising executive education. But we can do so much more to immerse the aspiring business cofounders in the startup culture! One way to do it is to run a Startups 101 course where a mix of VCs, founders, and developers will work with the students on specific projects and also to share their experience. 95% of all startups fail, and the experience of failure is invaluable. Learning from others’ mistakes can shorten a founder’s path to success dramatically, so this must be a course on failure as much as it will be a course on success.

Interestingly, on the way back, I got a call from a Harvard Business School MBA student, contemplating a social recruiting startup similar to TopProspect. He systematically asked me about what worked and what didn’t, and I was happy to share that knowledge which would possibly make him succeed. The mind economy of the Bay Area is an amazing culture, and we will see new kinds of companies founded by new kinds of MBAs – hopefully many of them from Wharton/West!

Jack London Marina

I’ve attended the first ever AWS Re:Invent in Las Vegas, November 27-29, 2012. It was great to reconnect with my old buddies from Amazon, including Jeff Bezos (if you include into buddies folks chatting in a PacMed cafeteria line). I was there when AWS started out in 2006, and a user of AWS – exclusively at TopProspect and a Spark cluster at Klout, and now running Versal fully on AWS (although properly abstracting the cloud provider).

The session on online education was one of a few unusual ones, devoted to an application area instead of an infrastructural issue.

The panel was run by AWS Director of Education, Steven Halliwell.

AWS in fact has a special Education group. Steve explained the subtitle of the session, “A Seismic Shift in Education,” as a source of the metaphors AWS uses consistently and for a reason.

The very existence of this role was encouraging. The panelists were: Dr. Anant Agarwal, the CEO of edX; Michael L. Chasen, a cofounder of Blackboard; and Prof. Bill Howe of University of Washington, a member of the EScience Institute there.

Some of the highlights. EdX is a non-profit, and Anant said they evaluated a ton of platforms before making their own. They wanted an open source, cloud-based one; none had all the desired features, although one was close, except for not being in the cloud (NB remember which one). The non-profitness and founding by Harvard and MIT are their strengths vs. e.g. Coursera – why would educators give their data to a commercial entity?

Blackboard still operates eight data centers around the world, but augments it with the cloud. Often their installs start under an wary adopter professor’s desk, and then run hundreds of courses before Blackboard gets a call.

Bill systematically presented data mining revolution in science as driving science and online education shift as well.

I’ve gained a lot of insights talking to the panelists after the panel, and also asking a question in session about MOOCs scaling the wrong part of the university. EdX is trying to be more interactive and compliments the original live courses with the X version of online courses; but no comprehensive answers were available, befitting the complexity of the problem (to which I have some nascent solutions.)

I’ve been a Meetup.com organizer for more than a year, setting up Scala for Startups and co-organizing Spark Users which started as one of the early Scala for Startups meetups.

The third meet up which I setup up, The Rank Gang, was in the works. It turns our that edtech industry needs trust most of all – as recent Coursera cheating events demonstrate. I was planning an edtech meet up with clear engineering focus for a bit. Since an organizer is limited to three meet ups, I’ve faced a choice to get a new one, or reorganize the Rank Gang. At it turns out extremely fitting that the trust and ranking technologies are now coming to the fore of the tech in edtech.

We will not limit ourselves to the trust and ranking issues only – the whole edtech area will be represented, starting with the state of the art web and app dev technologies and standards being adopted by the practitioners. Join the Rank Gang and see what’s next in edtech!

My founding message to the meetup with the new focus follows.

Dear Rank Gang Members – as some of you know, I founded this meetup based on the original Open Ranking Initiative, set in motion in 2008-2009, as a result of my pioneering Ph.D. thesis and web-scale data mining research on communications ranking. I’ve set up the meetup when working on influence ranking in industry. However, influence ranking industry in its narrow sense proved not conducive to a scientific approach nor was it fostering the community around the ranking technology.

The reason for this is that the whole foundation of influence ranking is not at the point where it can be benchmarked safely for the businesses promoting it. Influence ranking stays elusive – as Bakshy, Watts, et al submitted, in their Twitter analysis entitled “Everybody is an Influencer,” it is extremely hard to come up with a valid model distinguishing actual generators of long-range influence cascades. I believe that real influence ranking is inseparable from message modeling, not just “influencer” modeling. I’ve discussed it with Profs. Yuri Leskovec of Stanford and Lyle Ungar of UPenn as a dual problem – you have to model both viral messages and viral “influencers” together. Is it the message or the messenger which carries “influence”?

Real answers will not come from any single company. Formalizing the Open Ranking Initiative, I’ve proposed Open Influence Exchange as a platform for comparing multiple rankings from multiple vendors. Furthermore, I’ve become convinced that “influencers” and their ranking is an artificially narrow endeavor. True thought leaders are not always considering themselves as “influencers.” They can be teachers; experts; curators; doers. Brands seek customers, and all of the above can be their trusted intermediaries. We need a much more nuanced, finer-grained system of ranking people along with their knowledge and interests. And it all starts with trust. Any system ranking people must be verifiable and trusted by the general public.

Now, with the co-founding of Versal, I am working on a global platform addressing key problems of online education. Trusted ranking is one of such problems. In education, ranking starts with grading, and finally leads to a priorities list of prospects hired by employers. Everything in that setting should be enveloped in trust. The practice is much better understood than influence ranking.

So this meetup will address ranking in the context of educational technology and the trust system it needs. We will cover everything edtech from the engineering point of view, while retaining a keen interest in trust and ethical people ranking. In the coming months we will grow our membership, including the key players in edtech. Hopefully your original interest will be fueled by this new focus. Please let me know if you want to help with the meetup and/or have ideas on the venues, speakers, and members to invite!

Cheers,

Alexy