Billet Night at the Coast Guard Academy this past March was full of the usual excitement. Senior cadets nervously clambered onstage as their names were called. Some jumped in the air upon learning their first assignments. Others squealed and. hugged new crewmates before posing for pictures.
You’d never guess that behind the scenes, there was a big change. This year, Officer Personnel Management (OPM-2) found a way to automate the assignment process. So instead of four officers spending a week manually making assignments one by one, an algorithm matched these cadets to their first jobs – in seconds.
“I think the team was pleasantly surprised at how effective the matching software was,” said Lt Cmdr. Luke Carani, who developed it. After running the program, which was written in Python, he and the assignment officers (AOs) verified the results. “We reviewed the final slate in detail to validate that the software program didn’t violate any constraints, and then the team certified the assignment recommendations with no changes. It was an optimal matching.”
Carani joined OPM-2 as its first software developer last spring. The mission was to determine how the office could automate and improve manually intensive processes. He ultimately set his sights on CGA and OCS first assignments, where class rank and few additional variables make matching straight forward. “Figuring out who turn it is to pick is easy,” he said. All cadets submit e-resumes where they rank their choices. Carani also factored in dynamic cutter berthing constraints, programmatic eligibility criteria, selection panel results (for flight and cyber), and “co-location” requests to ensure that the needs and desires of members, units, and programs were represented appropriately within the assignment process.
It took a couple of weeks to develop the program, which the AO team hoped would provide an 80 percent solution. “I thought it would help us improve the process,” Carani. said. But when OPM used it to assign all 259 cadets in January, it exceeded their expectations because it considered all assignment factors. The software program he created to place OCS candidates was similar, with adjustments for constraints like Special Needs and co-location, since officer candidates are often married to other service members. OPM used it to assign all 76 officer candidates in February.
The most common question Carani gets is whether using an algorithm, as opposed to having a person make each decision, is fair. “We just implemented in code what used to be done manually,” he said. “People are still getting their best-case scenario in equitable fashion. Before approving the assignments, OPM applied the same rigor to certifying the software recommendations as they do to all PCS assignments, and no changes were necessary.”
First Clas Cadet Kira Howard, a senior cadet from Wilton, CT, who majored in marine environmental health science, said she thought it was both fair and “pretty cool” that OPM was able to automate the assignment process. “I’m going to ATON (Aids to Navigation) team in Astoria, Oregon,” she said. “It was my second choice. My first was Hawaii, but I didn’t expect to get that with my class rank of 60. I’m extremely pleased.”
In addition to saving time, Carani says one of the biggest benefits of using the algorithm is the ability to simulate different outcomes. In the past, if a ship changed its capacity for new personnel, for example, the AOs would have to manually re-adjudicate that assignment and all others that followed. Now, with one change in assignment, others are reassigned automatically. The AOs can also explore different scenarios and see how that might affect assignments.
Carani’s next goal is to create a program to help accelerate the assignment process for the rest of the officer corps, and eventually enlisted personnel, too.
“The real benefit is that it frees up the assignment officers who do the nitty gritty work,” he said. “It frees them up to do the high-level career management people need to help them with their career progression. Let computers do what computers are good at, and let humans do what humans are good at.”
-USCG-