What do Optimal Stopping Algorithms have to do with the current H-1B landscape? Nothing. But what if Trump changed the method used in awarding H-1B visas from a lottery system to one where the highest pay gets filled? Then, it has everything to do with Optimal Stopping Algorithms.
As you know, I’m an attorney that likes to develop software. When I discovered the H-1B visa, I saw a chance to code something that blends my ability to practice law with my love of development. Work Visas was born. Then Donald Trump got elected. As of yet, he has not shaken up the H-1B process. I do not anticipate that he will do so this year because it is very close to the April 1st window to get your application into the lottery.
One thing Trump could change is the current regulation (not a law passed by Congress) that has been in use since 2004. This is the lottery that determines the H-1B visas subject to the annual cap of 85,000. Currently, only 8 companies scoop up 60% of these visas. Their average pay is $70,000 per H-1B, whereas Amazon pays $113,000!!!!!
What if Trump said that the applications will be arranged in the order of their salaries. What will the Indian IT firms do? They will very quickly be taking about 60% of the H-1Bs again if they want. First, they must apply Optimal Stopping Algorithms to fix the salary for all of their new H-1B applications. To determine what one works best, the data set of the 236,000 applications needs to be accessed to see what the average price of the H-1B visa is for various subsections.
The H-1B cap exempt visas have to be excluded, except for the 20,000 for masters or higher education. These exempt applications have their own lottery system. The 65,000 subject to the cap are all for bachelor’s degrees, so that pool may pay less. The average salary of these two pools will give a rough estimate what how much you should never go below, or could just be garbage output based on faulty inputs.
Currently the H-1B visas are awarded on a randomized lottery, at least the 85,000 cap subject ones are. The bids for the H-1Bs may not accurately reflect the average salary. The average salary could be higher or lower. To figure this, you can take a sample of years of data to average the randomized lottery to get a more precise idea of the average H-B pay. You you could get all the LCAs and rank them in order.
Optimal Stopping Algorithms need data
In fact, several websites do just this. They pull LCA data from the Department of Labor and push it into tables for you to review. Here’s the highest paid employers. They also provide the list of the pay for various professions and states with the most LCAs. This is somewhat misleading because an LCA is not an H-1B job. But, we’d know what level to bid for an employee to increase your chance of getting a job.
We need the data of the actual applications, which are not published from as far as I can tell. However, we can take the LCA data and cleave it in half, then add $100.00 to be safe. Of course, this fails because of the sample size. There were 236,000 applications. You cannot just cut it in half and add a bit of money. You need to get the top 85,000 salaries and add money to that.
The algorithm needs to find the salary for the 85,000th highest H-1B application, and add ten bucks or so to it. Then, you’d start bidding up the price as the year goes on. But at least you would know where to begin your application to ensure a visa. That is, you would, IF Donald Trump changes the ways H-1Bs are awarded from randomized to based upon salary.