"...people and computers can be connected so that – collectively – they act more intelligently than any person, group, or computer has ever done before."
- MIT Center For Collective Intelligence
Hello, my name is Ron Itelman, and my passion is creating systems that leverage human and machine learning to augment efficiency and give users delightful experiences.
I have worn many hats and consider myself a generalist. I have had the roles of product designer, product owner, UX designer, full-stack developer, startup-founder, business manager, and most recently I have been studying neural networks and machine learning.
My specialty is working with data scientists, machine learning engineers, organizational leaders, and end-users to increase productivity while giving users experiences that feel magical. I don't use the word 'magical' lightly. I have been fortunate enough to hear feedback from co-workers that the innovation work I've delivered is magical.
If you want to transform the impossible into possible, let's work together.
THOUGHT LEADERSHIP ON DATA PRODUCT MANAGEMENT
Collective Intelligence For Government-Scale Analytics
One of the challenges in the public sector is being able to measure the impact of investments and share data in a universal format. Lumina was a concept that came out of an innovation effort and eventually led to work with the Denver Chamber Of Commerce.
The Myth Of Being "Data-Driven"
Unfortunately one of the challenges I've seen is that many organizations believe they are using data effectively because they generate lots of it. Quality often beats quantity.
On Converting Thoughts To Data
I've been getting more and more comfortable with regularly posting on LinkedIn and engaging with the community. I've learned a ton, met a lot of interesting people, and like to write about data product management topics.
The Business Intelligence Pyramid & Semantics
While researching the topic of why organizations aren't getting the value they expect out of their BI investments, we found the lack of an understanding in a semantic layer is a primary cause that prevents organizational leaders from achieving the results they want.
On Data Experiences
There are a lot of complex tools for semantic management. My focus is on simplifying complexity so that the worlds of data, design, and semantics can be on the same page to make effective products.
ON LEADERSHIP, CULTURE, DATA, AND PRODUCT MANAGEMENT
When we aren't in the same physical location and don't have whiteboards, I find that illustration is a great tool for communication, especially for abstract concepts related to data. Below are some illustrations as an example of visual communication.
I describe data management as the way companies can get teams cross-functionally aligned on concepts (semantics), the strategy to govern data, and the execution of those plans. The act of creating a semantic management layer is the same as a company creating a star chart that puts the sun in the center of our solar system, rather than the earth. Without semantic management, people operate in silos. By creating a map that puts semantics at the center of the entire organization rather than their own silo, everyone can have a common understanding of how their processes fit into a universal framework.
The most efficient teams are teams that can work in a synchronized fashion. Therefore, and effective product management organization is one that gets everyone on the same page, with efforts aligned. Focusing on alignment first before any particular efforts in any initiatives, is key to having a successful product management team.
Leadership & Culture
My leadership style is by aligning teams toward a common objective, while investing in understanding how people feel the most connected and most impactful to supporting the team and achieving the goal. By aligning people's intrinsic motivations and allowing them to express their creativity and how they want to contribute it creates a culture of inspiration and showing up with authenticity to give one's best.
Data Science For Sales Training Meets UX Research & Design
I was brought on to PagerDuty to build a Tribal Knowledge Recommendation Engine using AI. After analyzing PagerDuty's operational processes and sales training systems I was able to identify a key high-value problem: The demands on content teams and sales teams for both sales training and client preparation were exploding as the sales team grew. Due to the recent IPO, there was massive pressure to rapidly scale the sales team to meet forecasts and expectations.
The simple solution I worked out is to break a Google Slides or Office pitch deck down into JSON and labels using a database schema that could resolve situations to individual content items such as images, words, etc., using standard machine learning techniques.
Labelling data is a key ingredient of how to program machine learning systems. Therefore, one of my goals was to figure out how to label the content within each slide of their pitch decks and map them to their sales training system. By doing this, sales and marketing can share real-time business concept analytics to see what is working, and what isn't.
“Success is no accident. It is hard work, perseverance, learning, studying, sacrifice and most of all, love of what you are doing or learning to do.”
Innovation In Education: AI, Emerging Technologies, IoT
I joined Pearson in late 2016, and I am lucky enough to have worked on innovation efforts for some of Pearson's biggest and most high-profile projects. This included the IBM Watson partnership focusing on AI for teachers and students, as well as our R&D team's innovation work on Pearson's flagship college math product. Most recently I have been working on innovation projects involving AI, IoT, and future user interfaces.
LAIA (Learning AI Assistant) was an emerging technologies / innovation project that included using projected light to turn any desk into a touchable user interface. Starting with foam and a lamp, we were able to rapidly iterate and work on 3D printed designs to bring an affordable AI tutor into every student's home.
One problem that many organizations face is that teams work in silos and there is no easy way to synthesize prior research across various departments. Using TensorFlow, I aim to use machine learning on our existing research to power a search engine. I use D3 to create Hierarchical Edge Bundling visualizations.
The challenge for this project was to fit a math palette, learner history, math content, and learning aids all on a responsive mobile device experience. Most of Pearson's projects are designed for desktop first, and I was able to achieve a beautiful and elegant mobile-first design that is fully responsive with micro-animations. After an intense brainstorming workshop I worked with our machine learning experts and data scientists to design the user experience and then coded the functional prototype. The project was heavily focused on leveraging machine learning to dynamically predict student mastery at a fidelity the company had previously not achieved. This project included handwriting recognition, and integrated various open source libraries which opened the door for dynamically generating personalized user experiences on the front-end.
One of the challenges in using machine learning for adaptivity is that it is hard for a non-data scientist to understand what is happening inside the 'black box' of the algorithms. To help business leaders, learning scientists, and data scientists 'see' inside the black box, we created a real-time dashboard so that a viewer could observe an individual student's mastery levels in relation to knowledge maps and expression trees.
Early Alerts is a product focused on helping teachers identify who is struggling, where they are struggling, and what interventions they can take to prevent students from struggling before it gets too late. Early Alerts is a data visualization tool powered by the amazing data scientists at Pearson. Early Alerts is undergoing beta testing and is slowly being rolled out to universities across America and eventually the globe.
Part of the user experience with math, let's face it, is that it is not always fun. Yeti's Peak was a rapidly developed mobile game prototype, based on DuoLingo, to give students simple math concepts that they could do anywhere, and anytime, to help sharpen basic math skills.
“The best way to predict the future is to create it.”
- Abraham Lincoln
Social-Emotional Analytics For Education
In 2016 I entered the Startup Weekend EDU hackathon, where I met some fantastic people passionate about innovating in education. At the event participants got on stage and pitched their ideas. Mine happened to be focused on capturing emotional data from students in order to provide teachers with analytics correlating emotion to student performance and identify behavioral issues early on that affect classroom management. Additionally, to address the problems of student bullying and other challenges students might face, I hoped that students had an easy and private way to reach out for help. I was very lucky to have several exceptionally talented people join my team, and we won 1st place!
The signup page was intended to be fun, and uses css animations to move the clouds across the screen as well as rotate the sun and its rays of light. I designed and coded both the front and back-end using Firebase.
After the hackathon we piloted Happy in a school and were able to demonstrate that students enjoyed logging in their emotions, and that teachers and school administrators liked having a dashboard where they could see how their students were doing. We provided a series of filters and warnings that a school administrator could use to preemptively address potential issues. Synthetic data shown.
The user input from students was intended to be as quick and easy as possible. A student would select an emoticon on their device or the classroom's shared computer, with security and privacy in mind. This data generated a labelled data set that we could correlate to student behavior and performance, as well as teacher interventions. With that data we could then use machine learning to predict which students might need help before behavioral issues arise which can disrupt the entire class, a major problem for teachers.
“If you think good design is expensive, you should look at the cost of bad design.”
- Dr. Ralf Speth, CEO Jaguar
Next-Generation Data Visualization and Machine Learning
IBM Watson Health acquired Truven Health Analytics, a leading provider of cloud-based healthcare data, analytics and insights in 2016. I was brought on to help with a couple of specific projects via a short-term contract. At IBM Watson I worked on a data visualization engine concept that would allow users to select the best visualizations according to the types of data and business questions they had. Additionally, I utilized IBM Watson's cognitive capabilities to allow the UX Research team to extract concepts, keywords, and sentiment based on user surveys. I also worked on telling stories through data.
This data visualization engine concept allows the user to visualize information in different designs and select which is most effective. On the left the user can see the data's story, get information about the visualization, the pros and cons of using it, experiment with design, engage a 'What if?' engine, make annotations, identify key insights, discuss the visualization with their team, and access the data and code from one dashboard.
Utilizing Watson's cognitive capabilities I was able to parse through the UX Research team's user surveys to generate an understanding of common and searchable themes, as well as immediately getting a snapshot of whether user input is generally positive or negative.
I was tasked with telling the story of the UX Research team's data and findings for a large poster (42 x 90 inches). This persuasive design visualization shows the timeline, methodologies and participant information, as well as the key insights derived from that research.
“Simple is hard. Easy is harder. Invisible is hardest.”
- Jean-Louis Gassée, former Apple executive
Re-imagining A Patient Management System
I joined ClearChoice, an oral surgery and dental implant center, with the task of re-imagining their Patient Management System. I relished the challenge brought up during my interview by the VP of IT when I said I was interested in learning systems '...if you can find a way to build intelligence into this that would be great'.
I began my research by interviewing doctors, surgeons, front-office administrators, heads of departments, and other staff. To solve their problems of conflicts with doctor scheduling, I first focused on understanding the building blocks that were being utilized by individuals and departments. I mapped the relationships of their clinical and procedural data, designed the logic and user interface for a smart scheduling system.
The new system shows doctor times, is color coded to intuitively show information, is responsively designed, predicts conflicts, and is designed so that future versions will get smarter over time.
Patient information is also captured in an activity timeline, so that with a quick glance a care provider can get an overall story of the patient's history from various data points, without having to scroll to different sections.
This digital periodontal chart allows users to rapidly capture gum and tooth health with hot keys. All the teeth and dental symbols were hand drawn. Additionally measurements are color coded to help give doctors and surgeons information intuitively.
"If a picture is worth 1000 words, a prototype is worth 1000 meetings."
– Tom & David Kelley, Creative Brothers at IDEO
Collaborative Video and Music
In 2012 I decided to take the entrepreneurial leap and work in another area that I am passionate about: music. Real-time collaboration was not possible due to latency issues with data transmission speeds. Additionally, in audio programs, looping is a superior way to practice and create music, however no video looping solution existed that synchronized video with beats per minute.
JAMR was a mobile app that lets people easily collaborate on music asynchronously with zero latency. Video frames and beats per minute were synchronized so that video loops repeat in tempo. This technology was successfully produced and a utility patent was issued in 2014.
A user records herself while another video plays simultaneously, the two videos are merged together into one.
A user clips a section of their video to loop a music phrase which gets shared online for others to download or used in video mashups.
"As far as the customer is concerned, the interface is the product."
- Jef Raskin, started the Macintosh project for Apple
Automating Style Guide Generation
Style guide driven development is a great way to organize design elements and front-end code, yet creating and maintaining style guides can be time-consuming. Additionally, css-conflicts and changes can result in lost productivity. In 2015 I created STYLI as a fun side project: an automated way for organizations to create dynamic style guides and shareable code components for designers and front-end developers. Users can drag and drop their html, css, and js into STYLI, and share their code components privately with their development team or with the public. STYLI has been described as a 'GitHub for design'.
Visually I used images related to verbs ('sign up'/'go up'), and style guide pages were kept in grayscale so that the focus was on the user's designs.
A user can drag and drop their html, css, and js into STYLI, and code components are dynamically generated and shared privately or publicly. Additionally, one can download publicly shared code.
Design and UX elements are organized into customizable groups. A user clicks on an element in order to get any css, js, and html related to the code/design component.
1st Place, StartupEDU
AI, Analytics, and Design
Biometric & Wearables
Rapid Product Testing
Audio, Video, Content Authoring
AI, Emotions, The Brain, And User Experiences
Currently I have been very interested in the concept of neuromorphism, the intersection of artificial intelligence, biological intelligence, data science, and user experience design. I recently gave a talk in San Francisco at the Data Visualization Summit (2018) on Human-Machine Cognitive Experience Design. The HMCXD talk focused on learning science, machine learning, and user experience design. Isomorphism is a general concept that appears in several areas of mathematics and software design. The word derives from the Greek iso, meaning "equal," and morphosis, meaning "to form" or "to shape". Neuromorphism is the art of designing intelligent systems that can be applied to humans and machines learning together.
The talk on HMCXD was an interactive session with the audience, where I demonstrated certain concepts live. I became obsessed with the questions of "How do you measure the intelligence of a system where humans and AI cooperate" and "How can you intentionally design and increase intelligence in a system"?
Yes, that is a T-Rex
Before the presentation started we had a T-Rex walk through the room in order to capture a novelty event in the audience's experience. Novelty releases dopamine in the brain, which is correlated with better learning impressions.
Cognitive Persona Archetypes
I became very curious as to what kind of cognitive experiences could be generated in order to achieve goals and objectives. I explored dynamically generating cognitive experiences (dopamine or serotonin, etc.), as well as exploring when users want a 'flow' experience vs. an 'efficient' experience. Of course, we had to take a photo of the T-Rex by the Golden Gate Bridge as a souvenir.