Six Coding Challenges

The work does not run in the order a normal project runs.

Research does not finish before design starts, and design does not finish before development starts. All three run at once, on the same few hours. A static comp does not survive a room that wants to see something work, and a developer sitting idle for an hour is an hour the team does not get back.

The challenge is not to make every screen perfect. It is to get the highest-value flow moving early enough that the team can test it, change it, and still have something credible to demo when the clock reaches zero.

Build, verify, and revise on the same clock.

The pattern held across every challenge. Ship a thin slice first: the first hour goes to one flow, built well enough that a developer can start on it immediately. Keep researching while the build runs: stress-test the flow against records with no owner, permissions a supervisor has that an employee does not, and edge cases that the business case may or may not actually contain.

When research changes the plan, help the room change it fast. A found edge case is not useful if it becomes a note for later. On a deliverable measured in hours, catching a problem and fixing it need to happen in the same conversation, which means bringing the team a specific change to make rather than a general concern to worry about.

That loop repeats, tightening, until the clock runs out.

The version with a full team behind it.

USCIS ran the RMSS Human Capital Management Solution Challenge on February 27, 2022. Ten people, eight hours, and a business case built around onboarding employees and handling their documents securely. I was the lead UX designer: personas, journey maps, high-fidelity Figma, and enough HTML and CSS to keep the front end from drifting away from the design mid-build.

Four personas got built that day: an HR employee who uploads a new hire’s documents but loses access to them the moment intake is done, an HR supervisor who is the only role that can reopen and redact a file afterward, a records manager who owns compliance without owning intake, and a security analyst who ran out of runway before the team could finish her. Her page still has the placeholder text nobody replaced. It stayed in the deliverable exactly like that, because that omission is more honest than a fourth persona finished after the fact would be.

The team’s own definitions page set the rule that shaped everything else downstream: a record is a document with a retention policy, and once something is declared a record, it does not get edited, it gets a new version. That single sentence is why the redaction flow works the way it does.

A supervisor opening a passport image to black out a Social Security number is not editing the file. She is producing a second, separate version next to the first, with the original left untouched underneath it. Which role can even reach that toolbar was decided on the spot: full access for the supervisor, none for the employee once a document is filed, and a records manager who can confirm a file exists without ever seeing its contents.

The reporting tool is the part of RMSS that reads as the least like an eight-hour build. It shipped two ways to ask the same question: a drop-down query builder for a records manager who has never written a line of SQL, and a raw SQL editor for whoever the technical member of the team was, both feeding the same results table so neither path was treated as the throwaway option.

It also shipped the actual failure states: an empty results panel, a loading spinner, and a required-field error on the rule builder, not just the version where everything works.

SELECT name, size, sum(number) AS total
FROM staff-2014
GROUP BY name, size
ORDER BY total DESC
LIMIT 10

Not everything got the same amount of attention. The sign-in flow, password reset screens, and base employee-profile form kept the placeholder brand of the component kit the team started from rather than being reskinned. On an eight-hour clock, that is a deliberate trade, not an oversight: time went to the parts specific to USCIS and human capital management, not to screens every enterprise product needs and few evaluators score closely.

The muscle has to stay sharp between live challenges.

Separately, and on my own initiative, I built a second exercise in Adobe XD: a five-hour, one-person build of an Employee Self Service tool covering timesheets against project charge codes, overtime requests, and absence requests. It was never submitted anywhere. It exists because the gap between live challenges is exactly when the muscle atrophies, and a timed solo drill is a cheap way to keep it from doing so.

Smaller in scope than RMSS by design; one person cannot replicate what ten people cover in the same span. What is consistent between the two is the discipline: build the flow that matters first, and do not let a compressed timeline become an excuse to skip a state a real user would actually hit, like a day with a mismatched hour total or a charge code someone forgot to add before submitting.

Six challenges, six different results.

Naming all of them is more convincing than naming only the win. COMET is the best documented of the two wins: the GSA Blanket Purchase Agreement for CIO Modernization and Enterprise Transformation. GSA selected REI as one of twelve companies out of thirty-nine quotes submitted. I was the sole designer on REI’s submission for that challenge.

The honest version of that sentence stops there. The vehicle’s funding belongs to the vehicle, not to any one designer’s screens, and REI’s win belongs to the whole team that built and presented the submission. ODOS III was the second win, awarded to the ADG-REI joint venture in 2022.

RMSS is the one loss with a clear, documented cause. The challenge did not come down to the design work shown earlier in this study; an environment issue, specifically a technical problem with the AWS deployment, got in the way during the challenge itself. SPEED and FINCH did not result in an REI award either, with no design work surviving from either to show here. The remaining engagement, an earlier ODOS challenge, is still being checked against employment dates before it gets a firm place in this list.

One precision matters across the whole record: COMET is CIO Modernization and Enterprise Transformation, not “Challenge to Modernize IT.” “Sole designer on REI’s COMET challenge submission” is accurate; it describes who built REI’s entry, not who decided the government’s award. Nothing from before April 2019 belongs in this study, because that is when my REI tenure started.

Archive the work before the fire is out.

Neither of these should have needed reconstructing years later. The COMET challenge has no surviving screens at all, only the story of it and the award that followed, and RMSS survived only because one export sat untouched in a folder. Nothing about a timed challenge prevents a team from spending the last five minutes exporting the file somewhere it will still exist in three years. That habit is worth building deliberately, because the next fire always feels more urgent than the archive until the archive is the only thing that would have made the next portfolio update possible.

The unfinished fourth persona is the other one worth sitting with. It was the right call in the room: an eight-hour clock has to spend its last minutes on the system rather than on a document nobody would open that day. But nothing stopped a return trip afterward to close it out once the pressure was off, and that follow-up never happened.