Five Reasons to Work for the Department of Defense as a Data Scientist or Engineer
A series of Medium posts based on my book, “An Approach to Machine Learning in Cyber Defense for the DoD”.
Many in the Department of Defense (DoD) wonder what it is that might make an engineer or data scientist hang their hat at a specific location to work. In the study conducted for my book, I discovered five reasons engineers and/or data scientists would like to work for the DoD, despite all of the challenges I have highlighted in my previous posts on the subject (1, 2, 3).
Reason #1: The DoD has a ton of resources, financially and otherwise, to make your project happen.
The DoD has one of the largest budgets of any enterprise in the country when it comes to IT and tech related developments- and that is promised to continue to grow over the next decade. These resources and their subsequent allocations even down to a department level promise access to top performant platforms for technological developments, something that surely makes engineers and developers salivate. One of the delights of working at a large tech company per one of the technical interviewees for the study was the ability to spin up as much compute and storage needed anywhere in the world, at any time to complete projects. Imagine if the DoD could provide that to its engineers and technical corps? The potential and promise is there- the enterprise has collectively been moving towards this model ever so slowly but surely. If this could be provided to engineers, the promise of a DoD budget gives many an engineer hope that his or her project could surface into something tangible and potentially successful.
Another way these resources could be allocated is for financial incentives and bonuses for technical personnel. This is already being tried out here and there, but not with serious earnestness or at the pace the field demands in order for (let’s say) a green-suiter’s compensation to look close to what his or her income might be at a commercial entity. We have a long way to go to reach this point- and no, the promise of a sexy, exciting mission will not mitigate federal servant pay (despite what your leadership thinks…)
Reason #2 (and a popular myth)… Financial incentive and compensation.
A popular myth is that technical talent (engineers, developers, all types under the sun) receive better compensation at commercial companies than they might working for the DoD.
Truth is… if you’re a contractor, you’re most likely wrong!
Compensation in the $150k+ range for technical talent amongst DoD contractors is not uncommon at all… base salaries are usually pretty high for technical talent not in uniform (sorry, guys and gals in uniform- you’re just stuck).
So, lesson and myth buster #1- base salaries are oftentimes super competitive working in the DoD or intelligence community compared to the private sector.
Base salary is not the ONLY things your average tech worker is after- or at least this is not the full picture of what they are after. A mixture of benefits package (as a contractor, usually zero… not even health care) and a work/life balance as well as other compensation methods such as awarding stock are oftentimes reasons why many tech workers gravitate towards the private sector as opposed to the public sector.
One person interviewed for the study spoke to me about working on Wall Street, receiving a compensation somewhere around $600k annually as a data scientist working the stock floor… and it was miserable. Transferring to working for local government gave the data scientist a significantly lower base salary, but a much better work/life balance and less of a high stress position. These tradeoffs would not be that uncommon if those aspects of working for government were highlighted.
The oftentimes unfortunate truth behind working for the DoD, however, are things such as required deployments and an actually extremely unbalanced work/life scenario. There is a reason why the divorce rate in an elite enterprise such as JSOC toes the 80% range… work/life balance is not present at all. But is this something we have to subject our tech workers to? With the COVID-19 pandemic, remote work is becoming especially more prolific for those in the tech ranks, even in the DoD.
Do you need your data scientist out there right next to the warfighter? While that sounds romantic in theory, the actual truth is no, they oftentimes do not need to be out there with the cool guys doing cool guy things.
Actually- this generally will be a waste of time for the engineers, keeping them away from what they thought they signed themselves up for- engineering. A healthy dose of it is certainly welcomed, and some engineers might actually prefer the forever- deployed position more than working from home- here is a thought:
How bad would it be to allow the engineer and engineering teams to determine how often they need to deploy or be in the office in order to move their projects forward?
Giving them this type of responsibility and independence is what they crave (and frankly seek) when they put in applications to places like Google, Microsoft, etc… They’re most likely to make the most of the time they spend in high stress situations this way too. Rather than dragging their feet through a multi-month deployment living in a shipping container with less than optimal network speeds and compute, they can use their deployed time the way it should be spent- with the end user, in small, focused doses.
So- all in all, some of these financially associated incentives might actually be very easily addressed and provided to a potential tech worker looking at the DoD. The study (and myself, with many hours of thinking) did not come up with a solution around awarding stock to those participating in an organization… although there might be some curious work and study around running tech and innovation organizations within the DoD as startups, with a closed system that would determine the price of the “stocks” assigned to each organization, ultimately paid out as a bonus of sorts if someone sold the stock they owned in their respective organizations. Just- something to think about!
Reason #3: The DoD Mission
An interviewee very succinctly deemed it “The Coolest of Missions”:
“I’ve never found as pure a link between really meaningful outcomes and money… it’s an incentive that is almost impossible to replicate”
It goes without saying that many members of the technical corps would be willing to forfeit higher salaries for the promise of an exciting career at the DoD. Who wouldn’t want to work with spies? Who wouldn’t want to work with killer machines? Who wouldn’t want to be at the tip of the excitement, a front row seat to their favorite spy thrillers or war movies?
Please keep in mind- this argument for a “cool mission” only goes so far.
We have to be honest with ourselves- many in the technical corps experience what I call the “East vs West” battle- the one that goes back to the fact that the very polarized and vocal part of Silicon Valley at the moment (and for a hot minute) has never been on #teamDoD or really #teamamerica.
So, how can the DoD resolve this?
The DoD and intelligence community (whoo, scary that I mentioned them too!) must find ways to make war more accessible to its people- and therefore to the technical corps full of talent promise. War is as old as humankind, and while secrets seem super sexy, leaks are oh so common these days. The DoD and IC have a very bad tendency to not get ahead of these leaks with messaging- and do a very poor job at trying to explain WHY they continue to do what they do day in and day out, while those like myself who are still in the fight are reminded every day that there is an ultimate shining purpose, even despite our bludgeons.
So- messaging around the mission, how crucial it is, and ultimately WHAT it is- will have to be unlocked by the DoD community in order to attract talent at maybe slightly less desirable benefits package offerings, an absence of the promise of stock, and a sometimes very stressful work life that is not at all balanced with life at home.
Reason #4: The DoD is a treasure trove for data
The DoD has a ton of data…
This is way too oftentimes overlooked at a strategic or enterprise level by data scientists, data engineers, machine learning engineers and the like when considering job prospects. Let’s face it: the government has a ton of data. Because after over 20 years of constant war, you are bound to end up with — a ton of data to show for it, amongst other things. Way too many engineers do not know about this- and way too many DoD and IC recruitment campaigns miss the opportunity to tout just how much data they have available to experiment with, and just how many problems can be solved with that data.
Now- this narrative does not come without its own challenges. While the lack of data serves as a blocker for many nascent machine learning and AI projects data scientists are itching to get their hands on all over the world- historically, the attitude in the DoD has generally been one of negligence when it comes to the treasure troves of data built up over the course of several decades.
Who wants to be responsible for signing off on the storage of a ton of enterprise IT signals from your organization’s network that literally only still exists because auditing laws require an organization to store it for x number of years? The answer the DoD has made very clear in many instances is : nobody.
Nobody will champion on behalf of the unsung datasets that are collecting dust, siloed, classified at ungodly levels, or just plain ignored that present a challenge that the next brilliant engineer just would not be able to resist tackling. Yet another way the DoD and IC can mend its messaging campaigns- data scientists and engineers need and LOVE data.
So, sing about the data you have, even if you think the data types are unsexy or the mission they belong to is not the tip of the spear. Truth be told, most of those datasets are holding the keys to many, many extremely important enterprise challenges that are likely costing the defense enterprise millions of dollars each year..!
But what, one seasoned (likely GS-14 who is embittered about the fact that they have no foreseeable career promotion in the next… ever really) might ask:
“But what about the classified data? You know, the super top secret stuff that we can’t share with damn near anyone (let alone inter-agency)?”
This will have to be resolved in order to make the data meet the talent where it is today. The answer is certainly not lowering the level of vetting/rigour around our security clearance process (although it rarely seems to catch the popular insider threat types we have seen so far…) and the answer is not to hire a bunch of tech talent and keep them in the doldrums working on stupid unclassified projects that don’t matter for 2–3 years until their SECRET clearance comes through. There are plenty of techniques being worked on by the privacy field right now that would unlock these datasets, either in abstract or indirectly, to allow lower cleared personnel to start working on them NOW.
Techniques such as differential privacy (local and global techniques), federated learning, and others could be the key to not only unlock these datasets to talent waiting to tackle problems presented by the datasets- but even encourage more inter-agency collaboration without compromising SCI silos.
Are these techniques hard to implement? Yes, surely. But consider it an innovation bullet on your next performance report when you’re trying to go from that Step-8 GS 14 to — I don’t know- a Step-10 or whatever. I don’t know how those things work, to be honest.
Reason #5: The power of the top-down military chain of command to back your project.
In my other posts, I have oftentimes illustrated this authoritative, no-negotiating top-down chain of command that the military is most known for as a negative aspect of the DoD culture that has been fostered in the recent past.
However, in this case, there is one positive outcome from this totality of top-down command when you get the right leader in place who champions the power of technology and technologists against the mission, and when it is in place, it becomes extremely powerful.
An interviewee describes this power of command as the “hidden lever” to get so many high-risk and high-reward or promise technology projects going. A forceful and knowledgeable commander who empowers his or her technical experts can create and make way for projects that has never been seen (except maybe in a Musk company…). There is virtually no equivalent to this type of totality in command in the commercial world, where employees can quit at will if they do not agree with the mission his or her boss embodies.
When this power is used tactically and thoughtfully, this power of command can ben extremely enticing to engineers eager to serve an exciting mission in an organization where technology has been mandated as a top priority in a specific and tangible way against the mission.
Some last challenges to address…
All of the above might paint a fairly rosy picture when it comes to working for the DoD… aside from the articles referenced at the beginning of the post, there are some “smaller” challenges worth mentioning here that would be relatively easy to mitigate compared to some of the deeper rooted challenges such as those related to culture.
First- the “power of critical mass” is the attraction of like mindedness- engineers want to work with other engineers. And generally, they want to work with other engineers who are just as smart if not smarter than they are.
This is what one of the interviewees in the study called the power of critical mass- attracting the type of talent you want to hire by hiring more of them.
Gaining this type of attraction is hard enough for even large tech companies to accomplish, especially when it comes to the scramble for AI and ML related talent… but has proven as a powerful tool time and time again for engineers and PhD types.
Another challenge presented when working for the DoD or agency associated organizations is the ability to publish and share findings and your work. The ability to publish and share code, findings, tutorials, discoveries and creations is a huge part of what makes being an engineer fun. The ability to contribute to open source communities, either by starting one or contributing to a well established one, is also an enjoyment essential to many engineers’ identities that is often left in the dust when one works for the DoD due to reasons ranging from intellectual property stipulations for contractors and military personnel to data and classification sensitivity. Many interviewees, both technical and non-technical, expressed frustration at how difficult it is to publish while working for not just the DoD but any government body, hindering many PhD candidates and others from applying to government jobs at all.
The DoD (and other government bodies and agencies) must address and overcome its bureaucratic practices to enable fast and agile technical development of engineering solutions that can then be published by their engineers to a community. It doesn’t even have to be the public for some types of work that are especially sensitive- but fostering some type of way to publish works internally, share those publications at talks, or even foster inter-organizational or inter-community “government-sourced” areas to collaborate on technologies would help alleviate some of these pains, while giving engineers the satisfaction of being heard and sharing their findings. Hosting blog posting sites that are modern and appealing to write on (like Medium, not, say… IntelLink) would also help alleviate this pain a bit as well.