Taivara

Building the World’s Largest Performing Arts Ecosystem The Audition Tradition Performing arts programs in colleges and universities all have one thing in common: The audition. If you want to go to school to study the performing arts, you can bet that you’ll have to audition. For a long time, it didn’t matter where you lived or what your economic situation was. If you wanted to get in, you had to travel and take one shot at impressing the adjudicators. This was problematic for both parties. Travel was expensive and time-consuming for a 17/18 year-old just looking to go to college, and adjudicators spent days at a time sitting in an auditorium listening to auditions. Programs sometimes made attempts at digitizing the process. They started accepting CD and DVD recordings of auditions but soon found themselves buried under stacks of discs and mismatched paper applications. There was no escape from this annual headache. The Simple Solution A startup asked Taivara to build a solution. The idea was to take the audition process and put it online. Applicants could record themselves with a phone or video camera and upload the files to their online account, submitting them with their college application. Adjudicators could simply log in to review and rate the digital auditions on their own time. We helped them research the market to see if this was a viable product for the market to buy. Once they got commitments from customers we helped them raise enough capital to get the company rolling. Not-So Simple After All The founders had an MVP built by another firm. The MVP looked great but was technically weak. The product could barely handle more than a few users and didn’t take into account the requirements of the industry. We would need to build re-do the architecture to create a product that could handle the usage we expected. After digging deep into the needs of the industry, we started to realize how complex the product would need to be. To name a few complexities, applicants would need to use one account to apply to multiple universities, often using the same audio and video files for multiple applications. Programs would need to have different permission levels for different types of adjudicators and would often need one account to work with multiple programs. The web-app would need to process humungous video files and accept payments for application fees… No matter how complicated the requirements were, we were able to prioritize the necessities, roadmap out future feature sets, and ultimately build the technology to solve these complex issues. Caution: Heavy Traffic One of the unique aspects of the industry was the application deadline. Top-name universities often set their deadlines on the same day. Unfortunately, it was quite common for applicants to wait until a day or two, even a few hours, before the deadline to start their application. This meant thousands of users would be bombarding the servers, each uploading gigabytes of files, all at the same time. Taivara had to develop a web-app with enough server power to handle the heavy traffic while keeping costs efficient during the off-season. The Code To build out the back-end, we used a PHP framework called Laravel. This framework allowed us to easily cache ginormous media files for faster load times, send outbound emails seamlessly to our many users, write complex logic for our complex requirements, and fluently create and run database queries. The Art of Iteration After launching, we created tight feedback loops with key clients. Their feedback, along with our product knowledge, helped us quickly iterate and deliver necessary features, ensuring a healthy relationship with the clients that mattered most. The product continuously improved, allowing for more client acquisition and growth. Off to a Great Start The startup earned the trust of over 200 performing arts programs within the first two years of operation. The product designed and built by Taivara served as the primary digital audition system for some of the world’s most prestigious programs such as Juilliard and Cincinnati College Conservatory of Music. After about 19 months the startup had sufficient cash flow. We helped them source, interview, and onboard their own in-house development team. By the end of year three, the company had over 500 programs within their network and became the world’s largest performing arts ecosystem. Read More

Improving Customer Experience and the Bottom Line Out with the Old A Fortune 500 financial services firm came to us to solve a problem. Their customers were experiencing crazy high abandonment rates when signing customers up for their private-labeled retail credit card services. Whether at the register or online, too many customers weren’t completing registration. This problem was directly affecting their bottom line. The Challenge The idea was to make in-store and online registration a breeze. To do so, we needed to build a digital application that helps shoppers type less. The app needed to pre-populate as much of the application form as possible using only a few pieces of data. One example is for a customer could simply scan their driver’s license with the phone’s camera and most of the application would complete itself. The challenge… Design a visually appealing interface that could easily be white-labeled for each new brand. Structure a codebase flexible enough to support multiple data-collection workflows and brand-specific customizations. Support heavy traffic from several high profile brands. Meet aggressive development deadlines. Spotting Obstacles Too Many Keystrokes are Daunting In conducting early research, we realized that every keystroke brings users one step closer to abandonment. Long rows of inputs and legalese tend to be demoralizing to even the most determined of shoppers. Powerful Technology can be Slow Sure, we could implement the technology to save consumers a lot of keystrokes. Unfortunately, it could also be extremely slow to run. For example, extracting and formatting meaningful data from the photo of an ID can take as much as 10 seconds to complete. Wait times like these are almost as daunting as too many keystrokes. Allowing Various Workflows is Hard The beauty of this app is giving the users the power to choose the data collection workflow they want. However, each data-collection strategy requires a unique workflow with unique edge cases. In spite of this, we’d need to be able to seamlessly add and support any data-collection workflow into the app. Avoiding Obstacles With these issues in mind, we started sketching out possible user flows. After multiple user interviews and several iterations of sketches, we landed on an option that would eliminate slow loading times, minimize user keystrokes, and allow us to swap any data-collection strategy in or out with ease. With these workflows finalized and approved, we were able to move confidently along with the rest of the design and development work. Reacting with Rails We knew we had to choose technologies that would allow us to strike a balance between building quickly to meet deadlines and providing the smoothest, snappiest user experience possible. Rails allowed us to build out an API quickly and focus the bulk of our effort on the client-facing experience. React with Redux provided a strong technical foundation for the user interface that we can continue to build upon, regardless of what technology is ultimately used on the API side. In With the New We launched for a single brand with incredible results. We reduced the application abandonment rate by 87%. Success! The app landed our client 30 new accounts in one store in one day while significantly reducing the time to apply. Since then, we’ve matured the platform. We implemented new data collection strategies and have signed up numerous new brands. Read More

Mapping the Network Developing & Testing a New Data Visualization Approach Too Much Data The more we use technology the more data we create. In fact, every day we create over 2.5 quintillion bytes of data. At a glance, this sounds great. The more data we have about anything, the more we can understand and learn. However, industries are reaching a point where there is just too much data for them to analyze. Especially in cybersecurity. Old Visualizations In cyber-security, analysts look at all the different devices connected to their network in order to help identify threats and attacks. They look at when new devices join the network and how they are connected, along with many, many other pieces of data. This can be very difficult to keep track of when you’re watching over a large network, such as a large company. At the time, there were tools available to help visualize this data to make them more digestible. However, once the network reaches a certain size, the visualizations became unintelligible and ended up looking like a giant hairball. Yuck. Mapping the Solution Our client had done a lot of research into the space to identify why the current solutions were not up to par. Most solutions were taking a “force-direct graph” approach. Again, this worked well for visualizing small networks, but as the network gets bigger, the visualization gets more unintelligible. The better solution was to take a “Circle Packing” approach. This design allowed for a clearer understanding of how data is organized as well as great interaction with thousands of data points through zooming, filtering, and highlighting. They had built a prototype using this new approach and, even though it was an improvement from other solutions, it still needed to be a turnkey solution for giant data sets. They needed a productized solution and asked Taivara to build it for them. Building through Experimentation Using D3.js we experimented with a few different layouts that retained relative node positions for our data sets. After getting close with a few, but not close enough, we built a custom algorithm to do the trick. The prototype had unknown devices (read: potential threats) orphaned in the diagram. At a large scale, these unknown devices were scattered throughout and unrecognizable. We played with a few options but decided on grouping them together so that defenders can find all threats in one place. We also tested various different zooming, panning, and filtering interactions to find the right balance of each to make navigating the diagram easy. To do all this we used a combination of D3.js, React, Redux, RxJS, and Ramda.js to make the UI clean and snappy. Result After launch, the product received great feedback from the cybersecurity industry including many experts in the field. We continue to update and improve the product as well as apply it to other industries outside of cybersecurity. Read More

Saving the World from Bio-Disaster Identifying and Preventing Harmful Genetic Sequencing Deadly Bio-Threats A lot of people don’t know, but there is an entire industry of synthesizing DNA sequences. With emerging technologies like CRISPR scientists can quickly edit and create custom DNA sequences. What was once only an idea in science-fiction films like Gattaca, gene-customization is now becoming a reality. Science is taking its first steps toward eradicating deadly diseases, correcting fatal genetic mutations in fetuses, and even resurrecting extinct species. The technology is becoming less expensive and widely available around the world. While this may seem like fantastic news, gene-editing can pose a serious threat to national security. Director of National Intelligence James Clapper has included genome editing as a potential weapon of mass destruction. In the wrong hands, bioterrorists could edit and combine less-dangerous genetic strains into extraordinarily powerful diseases, viruses, and neuroweapons that could wipe out millions. “Given the broad distribution, low cost, and accelerated pace of development of this dual-use technology, its deliberate or unintentional misuse might lead to far-reaching economic and national security implications” states James Clapper. For less than $500, someone could potentially order a few seemingly-ordinary genetic strains and combine them to create, for example, West Nile virus.   Commercializing IP A client of ours traditionally worked as a government contractor, bidding on and executing large government contracts. They wanted to try something different, however. They wanted to create and directly sell a product of their own built off intellectual property they had developed throughout the years, opening new sources of revenue and proving they could profit from their past research and development efforts. The Cure Our client had developed a proprietary algorithm that could read genetic sequences and determine if they could be used to create harmful bio-threats. The plan was to productize this algorithm and sell it to gene sequencing labs to ensure they weren’t accidentally supplying bio-terrorists and heedless scientists with the tools to create a bio-weapon of mass destruction. Teaching to Fish Our goal wasn’t to step in, do the work, then step out leaving our client helpless when future work needed to be done. Our goal was to teach our client modern-day best practices for launching user-friendly, scalable products so that by the time our work was done, they could remain self-sufficient and productive. User Friendly First, we took their proprietary algorithm and made it user-accessible. Using Redux and RxJS we created a frontend architecture. This UI allowed customers to interact with the algorithm on the web with buttons and clicks rather than the original “command-line style” you used to see in the ’90s. We established core design patterns so that future changes and additions to the UI could be quick and simple. Scalable DevOps Part of creating the product included ensuring it was fast and could scale. This means the product had to compute quickly and handle increases in usage without crashing. Given the intensity of this algorithm, we needed a lot of computing power for users to access and run, sometimes many at once, sometimes not. Using Docker containers in Kubernetes, we could quickly spin up more server space at the moment it was needed ensuring quick access and minimal crashes. When space was not needed, we could spin down servers saving money on server costs. Safe & Secure Obviously, a product of this nature needed tight security built in from the beginning. Besides just the usual data encryption, we built very stringent security measurements into the app. We hosted the application in AWS GovCloud which is designed specifically to address strict U.S. government security & compliance requirements. We also stored everything in various locked-down subnets that could not be accessible to the world wide web so that we could ensure only the right people could get access to the right networks. Once everything worked and security was tight, we were ready to launch. High Score The product received high attention and praise from the company’s CEO. This was a giant step toward new beginnings for the company. When the product was finalized, it underwent a certification test for DNA sequence analysis, the standard for the industry. The goal of the test is to take a large test data file and find as many threats as you can with a minimum amount of false-positives. Read More