Colorado is known primarily for its breathtaking wilderness, but the Centennial State has also contributed a fair share of inventions to the world. The root beer float, Crocs and Christmas lights were all invented here. In addition to tasty beverages, comfortable but questionably looking shoes and essential holiday decorations, ambitious Coloradans have also been behind quite a few technological breakthroughs. At the University of Colorado Boulder alone, 3D printing, lasers and TiVo were all developed. The University of Colorado system collectively came in at number 36 on the 2022 edition of the National Academy of Inventors list of the top 100 universities worldwide granted U.S. utility patents.
Today, innovative tech comes from both the state’s universities and the private sector. This breakthrough tech takes many forms, from boundary-pushing industrial 3D printing to robotic recycling technology and platforms that make it easier for sustainably minded companies to monitor and reduce their emissions.
Built In Colorado recently spoke to engineers working at these companies — along with those at other innovative firms — to learn more about the exciting projects they’re working on along with how the culture of their engineering teams enables them to tackle such complex challenges.
Flatfile’s data exchange platform enables developers to create APIs that pull and convert data from files. Artificial intelligence is the talk of tech at the moment, and Joel Grus, principal software engineer, said he recently worked on an AI assistant for Flatfile, with the project taking on a “true full-stack nature.”
What was a product or feature you recently helped develop at Flatfile? What does it do, and what was your role in creating it?
Flatfile is a data exchange platform for file-based data import. Customers set up and configure spaces where data owners can upload, clean, transform and deliver data. Most of the time, data owners are limited to manually modifying and cleaning their data in order to get it ready to upload, but with Flatfile, developers can implement data hooks to programmatically transform their data.
Another engineer on the team created a prototype text-to-transformation “AI assistant” feature; for example, “If the state is Texas, set the greeting to ‘Howdy,’ otherwise set the greeting to ‘Hello.’” I took the prototype and focused on productionizing it, adding AI assist to our web application, creating a new back-end job to apply the AI-generated functions, providing the assistant information about the dataset schema, implementing security and rate-limiting, and making it a smooth and delightful user experience.
Transforming someone’s data in response to a text command is in some ways a risky thing to do, so getting the UX right was an interesting challenge.”
How did your engineering team culture support the successful creation of Flatfile’s AI assistant?
One aspect I found personally fulfilling was the true full-stack nature of the project, which included adding new API methods, creating new abstractions for data transformations, introducing new React hooks to manage the AI assist requests and iterating on the actual front-end user experience. Transforming someone’s data in response to a text command is in some ways a risky thing to do, so getting the UX right was an interesting challenge.
The other exciting aspect was demoing the assistant for teammates and customers. “AI assist” is a hot thing right now, partly because it creates a cool user experience, and its results are interesting and exciting. It has been a lot of fun seeing what ideas people have for asking the AI assistant.
How did your engineering team culture support the successful creation of Flatfile’s AI assistant?
This was one of the first projects I worked on after joining the company, so in addition to understanding how the AI assist prototype was built, I also needed a lot of context about how the back-end API platform works, how the front-end user-facing app works, how our jobs system works and how our data model works. Because it’s a startup, everything is always moving fast, but everyone was excited to pitch in and help.
AMP builds artificial intelligence-powered software and robots to scale recycling operations and increase the recovery rate of recyclable materials. Nathaniel Been is a software engineer who has spent the past few years at the company developing a web-based data visualization platform called AMP Clarity. Been shared the challenges his team faced in developing Clarity along with how AMP’s engineering culture enabled him to turn a tricky customer request into a new feature.
What was a product or feature you recently helped develop at AMP? What does it do, and what was your role in creating it?
We build a lot of really exciting robots that are revolutionizing the recycling industry. But once a robot is installed and running, how do the customers know what their robot is doing? That’s where AMP Clarity comes in! AMP Clarity is a web-based data visualization platform built to offer users decisive data on how their robots are performing through both real-time alerting and long-term reporting. In addition, Clarity allows users to change the configuration of a robot on the fly, enabling them to get the best possible performance. I’ve been the primary engineer working on Clarity for the last couple of years, so I’ve been able to witness the platform grow from a few experimental pages into a fully featured application.
As a comparatively junior engineer, what I appreciate most is that I always feel that my options are respected and carefully considered.”
What was the most exciting or interesting aspect of working on AMP Clarity?
Working on any customer-facing data visualization platform, there’s an inherent struggle between creating a product that’s easy to use and providing the widest possible breadth of data from which to draw insights. You learn quickly that cramming every possible graph and metric onto a page may be a feast for the data-savvy analyst, but the average customer finds it overwhelming and unintelligible. As such, careful thought needs to be put into exactly what data we show and how we show it.
We had an old dashboard that contained a giant list of every error the robot had encountered that day, each one displayed in exhaustive detail. There was so much potential for data exploration, but we quickly learned that nobody had the patience to slowly dig through it, and this method of display lacked the ability to see at a glance how the robot was performing at a particular time. When we rebuilt this page, we instead displayed a simple color-coded timeline for each error, and despite the page having fewer details than before, the improved layout of the data drove significantly greater adoption.
How did your engineering team culture support the successful creation of AMP Clarity?
There are myriad aspects of AMP’s engineering function that I could laud here, from the express encouragement to take risks to the acceptance that failures will happen and are not shameful. As a comparatively junior engineer, what I appreciate most is that I always feel that my options are respected and carefully considered. Recently, a customer made a number of requests about new ways for Clarity to display data for some of their robots, changes that immediately concerned me as they were incredibly specific and would add a whole new page of graphs and charts that were very similar to existing ones.
There are other places I’ve worked where said concerns would not have mattered, where it was made clear that my role was simply to build what was asked. However, AMP is different. At AMP, I know that if I raise a concern it will be heard out, and so I designed an alternate implementation of the request, one which utilized existing assets and would be usable by our entire customer base. I presented this alternate design, and after some discussions and tweaking, it was accepted and has become the next exciting Clarity feature!
EnergyCAP’s enterprise resource planning platform gives corporate sustainability teams the ability to more easily track greenhouse gas emissions along with energy use and spending. Dave Mosemann, a product manager, developed the solution that enables companies to track their greenhouse gas emissions and said EnergyCAP’s collaborative culture was key to its creation.
What was a product or feature you recently helped develop at EnergyCAP? What does it do, and what was your role in creating it?
Over the last year-and-half we’ve been working on CarbonHub, our financial-grade carbon accounting solution that provides our customers the ability to track and report their Scope 1, 2 and 3 greenhouse gas emissions. The product provides the ability to record emissions for each source through manual entry, importing records in bulk or automatically by calculating utility bills and other financial data. It sources date-effective GHG factors from several libraries that are used to calculate emissions based on usage or cost of a particular resource and provides customers the ability to use their own custom factors based on unique scenarios. Customers can set targets and report to their organization and stakeholders on their progress over time through dashboards and reports.
As product manager, I guide our design and engineering team in the design and execution of the product and prioritize features that add to it. I also provide training on the product to other teams in our organization.
Promoting this level of [high] collaboration allows us to iterate effectively and deliver a refined product that brings tremendous value to our customers.”
What was the most exciting or interesting aspect of working on CarbonHub?
A core tenant of CarbonHub is the ability for customers to “log in and go” without requiring a major implementation project. This can be a challenging endeavor when your customer base is composed of large organizations and enterprises accustomed to having a tailor-made software solution. We spent a good bit of time in the design phase iterating on how to break down the tasks in the product into steps that would allow a new user to accomplish their goals shortly after logging in for the first time but also not slow down more seasoned power users. We also spent a good bit of time conducting comparative research to understand common industry terminology so that the terms and help text were familiar to our customers.
How did your engineering team culture support the successful creation of CarbonHub?
I work with a great team of designers and engineers who value collaboration, and it’s a big component of our company culture. The magic of working on a software project is the ability to iterate relatively inexpensively. For me, collaboration drives iteration. In all stages of the design and execution process, as the team collaborates we are able to share and challenge ideas, ask questions, share concerns and discuss implementation strategies that balance a good user experience while still creating something that is feasible on a deadline. Promoting this level of collaboration allows us to iterate effectively and deliver a refined product that brings tremendous value to our customers.
PAIRIN’s software is designed to make finding a job and charting a career path easier by bringing developmental programs and resources from government, educational and labor institutions together under one platform. Devin Owen, a staff software engineer, helped build the platform’s new “My Resume” feature, which uses AI to help job seekers tailor their resumes and cover letters. Owen noted that the project was started before ChatGPT brought large language models to the fore, meaning that PAIRIN had to build its own models.
What was a product or feature you recently helped develop at PAIRIN? What does it do, and what was your role in creating it?
PAIRIN helps job seekers by connecting them to careers, programs, services and jobs while giving them the tools to succeed along their journey. I’ve really enjoyed working on our new My Resume feature that will soon be added to the My Journey product. My Resume uses AI and natural language processing to help job seekers tailor their resume and cover letter for each job they apply to by choosing the most relevant information from their professional and education histories.
This allows users to move through the time-consuming parts of applying for jobs more quickly and ensures they are stronger candidates. We’ll be pairing this powerful technology with a job search platform that will help job seekers find the best fitting roles based on their skills and the direction they want to take their careers. As the primary engineer on the My Resume product, I worked closely with a team of developers and data scientists to create the language models, scoring algorithms and web interfaces that will be core to both of these new tools.
The most exciting part of building My Resume was integrating our NLP tools. We started work on this feature long before commercial APIs existed for large language models like GPT.”
What was the most exciting or interesting aspect of working on My Resume?
The most exciting part of building My Resume was integrating our NLP tools. We started work on this feature long before commercial APIs existed for large language models like GPT, so we needed to build our own models. After testing a few approaches, we found that customizing a more general language model to be especially attuned for the types of words and phrases in resumes and job postings was the right balance between high-level language comprehension and our domain-specific needs. Our most important insight, and the one that took the most discussion, was how to use such a powerful tool.
Ultimately, we decided that instead of having our language model write user’s content for them — which we believe is ultimately harmful to their career chances and to employers’ ability to truly understand their skills — our AI helps users find jobs that are a good fit and empowers them to choose what information they want to add to their resumes. The team showed patience and commitment by taking the time to test the models, ensuring we reduced bias and pushed for the most beneficial uses of the technology.
How did your engineering team culture support the successful creation of My Resume?
PAIRIN’s engineering culture is what all companies strive for but so few actually achieve: supportive of each other and our customers while pushing the boundaries of technology. There were easier ways to add a resume building tool into the existing My Journey product, but we pushed for a tool that would fundamentally improve the job seeking process for our users and make them more successful. We took the time necessary to build the best possible tool, even though it took longer.
One decision in particular that stands out to me was when members of the engineering and QA teams stood up for increasing accessibility so that the benefits could be made available to all our users. That thoughtfulness for our users pervades the PAIRIN culture and makes the company such a satisfying and impactful place to work.
Parsyl is an insurance company that leverages data gathered from IoT sensors to provide insurance for the shipment of perishable goods. Justin Lewis, a senior data analyst, said one challenge his team recently tackled involved building an algorithm to help fight a pervasive problem in the cold supply chain industry: false alarms.
What was a product or feature you recently helped develop at Parsyl? What does it do, and what was your role in creating it?
Parsyl customers use our IoT sensors to monitor the temperatures of shipped perishable goods and then analyze that data on our cloud-based platform. The data collected helps businesses make strategic decisions to reduce loss and waste across their supply chains. A frequent issue the industry battles are false alarms, which result when sensors are left on at the end of a shipment and removed from the temperature-controlled environment. This results in the recording of spurious temperatures and notifies users of alarms, even though there has been no actual temperature abuse of the product.
To prevent this, we released a house-built algorithm to identify false alarms and automatically adjust the data to trim the readings. This innovation gives users an accurate understanding of the temperature before devices are removed from shipments or coolers. It also helps users who are already resource-constrained block the noise and put their resources where they are needed most.
We released a house-built algorithm to identify false alarms and automatically adjust the data to trim the readings.”
What was the most exciting or interesting aspect of working on this algorithm?
False alarms are an industry-wide issue, so much so in fact that many companies hire dedicated resources to manually correct each clients’ shipment records. We were committed to developing a solution that addressed this issue but in a way that was not overly complex or opaque and didn’t require too many resources. It was a fun process to try and find the right balance.
The end result was a custom but relatively straightforward algorithm that reflected our team’s knowledge and experience with the data rather than adapting and tuning a standard machine learning algorithm.
How did your engineering team culture support the successful creation of this algorithm?
We were motivated to pursue a solution to this problem based on direct feedback from several clients who expressed frustration with false alarms that were distorting their data. An automated solution was necessary to allow these clients to get a true understanding of what their data was telling them and use that to inform strategic decisions. Developing, testing and implementing the solution involved close collaboration between data analytics, data science and software engineering teams. We did a lot of pair programming, iteration and kept the customers’ perspective top of mind throughout.
BAE Systems, Inc. builds space technology, including instruments, sensors and satellites. For decades now, the company has been using 3D printing — also known as additive manufacturing — to aid in the design and development process. That said, it recently increased its 3D printing capabilities by meeting a challenging client request, which David Waller, principal investigator for additive manufacturing, detailed to Built In Colorado.
What was a product or feature you recently helped develop at BAE Systems? What does it do, and what was your role in creating it?
Additive manufacturing, also known as 3D printing, is reshaping the production and supply chain ecosystem. Companies across a myriad of industries are exploring ways to leverage these evolving methods and technologies to develop designs that were not possible using traditional manufacturing methods. The aerospace and defense industry is among those embracing this innovative technology with diverse printable materials to increase speed and agility in the design and development process.
AM is the process of joining materials to make objects from 3D model data by successively depositing material in layers so that it takes a predesigned shape. BAE Systems began using 3D printing in the late 1990s to produce models, cable harness mockups and proof of concepts. Today, we are employing AM technologies in all facets of engineering, from rapid prototyping to tooling and ground support equipment to flight hardware.
The ability to rapidly 3D print these parts with advanced materials allows us to efficiently design innovative systems like satellites that could not be previously considered with traditional manufacturing processes.”
What is the most exciting or interesting aspect of working on additive manufacturing?
We have found that one of the most beneficial aspects of the AM process is that it mitigates complexity-related costs. In other words, 3D printing a more complex part with intricate design will not necessarily cost more than a simple design, with some exceptions. As with many aerospace companies, we produce complex components in limited quantities. The ability to rapidly 3D print these parts with advanced materials allows us to efficiently design innovative systems like satellites that could not be previously considered with traditional manufacturing processes.
How did your engineering team culture support the successful adoption of additive manufacturing?
AM is challenging traditional processes, and one example is large metal forgings. A recent customer had difficulty sourcing large titanium parts because the lead time for the rough shape surpassed 12 months, so we pitched AM as a rapid alternative. The customer gave us the opportunity to qualify flight-worthy parts with a build time of only 12 weeks. That exploration program was only feasible due to the engineering team and culture driving the AM ecosystem at BAE Systems.
Our culture is very collaborative and everyone searches for solutions to challenges. That pathfinder project has led to further opportunities and continues to stir requests from ancillary organizations facing the same challenges. Our knowledge base of additive technologies allows us to determine what AM-appropriate parts are and the most economic manufacturing method.
While traditional production methods will always play a role at BAE Systems, the use of AM will continue to expand as the technology evolves. We’re excited to be on the forefront of this evolution and contribute to its growth, as well as to the development of the next generation of skilled workers to enable it.
