physics vs engineering

04 Sep, 2025

so i'm an engineer now, apparently

So I'm an engineer now, apparently. A circuitous journey through a peripatetic life will bring you great variety. Q.v. my initial life plan-- study English Literature. Hence why I found myself at Trinity College, Cambridge, doing Natural Sciences, originally as a chemist. Then, I swapped to physics, briefly attempted to switch back, withdrew from the registration to continue on as a master's student, re-enrolled at the tail end of my Third Year, applied to several physics PhDs, got a couple, accepted one, considered rescinding my acceptance before ultimately taking it, attempted to leave, changed my mind and stayed, then decided to leave for real, before applying to... A PhD in engineering. At Manchester. Which I accepted, then declined, then re-accepted. Then from which I withdrew my final acceptance. Applied to Oxford, also for a PhD in engineering. Got in. Accepted the offer (that was two days ago). Wrote the blog post on convolutional denoising autoencoders. Wound up here.

Am I qualified to make consequential life decisions with far-reaching consequences? Certainly not for anyone else. I am myself barely attaining any form of lifetime linearity, meandering helplessly between various moments of indecision, attempts to reinvent myself, and plans to move city. All to end up (almost) in the same place it all started and where everything for me happened-- Oxford-- minus a couple of years in the '00s and then another few in the '20s when I was in the US and then the other place, respectively. I've lived here since I was eight, a level of stability that seems alien and unearned-- feelings most likely due to having moved around a lot as a kid.

Though I wouldn't characterise my upbringing as "nomadic", we certainly had our close brushes with the lifestyle as a family of three. When I was born at Huntington Memorial in the 2001 incarnation of Pasadena (where you could skateboard everywhere and everything was filmed in Hi8), we were living in a condo, which we promptly ditched in late '03 for my biological father's childhood home in Salem, Connecticut, which became legally condemned during our foray down the ol' paternal memory lane and had to be hot-swapped for an unimaginative but roomy 1930s build in "nearby" sleepy Hebron, where there was no lake but plenty of coyotes and a surprisingly pretty swamp. There was no litter, no lights at night other than the Moon and stars, and it was a place where you spent your days biking 'round the big concrete horseshoe of a road that encompassed your little slice of suburbia, which mostly went McCain in '08 (the Ramms being the Democratic radicals).

This may technically count as doxxing our successors at the homestead. But I've gone back to visit the place a few times since I moved out in 2010, and no one seems to live there. Plus I removed the exact address. All you have to go on is this big loop of asphalt, pictured above.

Then there was a cottage in Gozzard's Ford (just outside Abingdon, Oxfordshire) for about a week in the Summer of 2010. Then a pub in nearby Steventon for three weeks, following a couple of places in Abingdon and then Wantage, then a couple of depressing bedsits in Cambridge, followed by a pretty decent room in a very loud and noisy house in Cowley under the auspices of Linacre College, Oxford, where I pursued some graduate studies in physics, partially at CERN (ooooooh, ahhhhhhhh) where things move very very fast around big circles and, eventually, someone lets you take blurry images with very noisy cameras.

Makes Things vs Breaks Things

A very significant presence in my life for a long time was an engineer. Growing up exposed to the practice of engineering, you develop a tolerance to weird ideas and strange sentences composed of seemingly invented words, such as the perplexing "incandescence during adiabatic compression of air upon re-entry", or "determination of practical lower limits on frequency and lower-frequency acceleration amplitude for ordnance-excited pyrotechnic shock testing"¹. Also learned is a general insouciance to the look of the command line, or an IDE, or C++ source code, etc. etc.

Perhaps there are challenges to growing up surrounded by a moderately-famous and, in many ways, wildly-successful engineer, in addition to a formerly-acclaimed professional cellist and graduate of the UK's premier musical conservatoire. Especially if you, yourself, are a semi-decent musician and/or talented mathematician/engineer/physicist et al. who releases music and gets the occasional Olympiad certificate. Perhaps it gets even weirder if, instead, your big passion is actually Jane Austen novels and The Bell Jar, and you turned out a bit non-heterosexual, and own the most makeup of all your parents' friends' sons. Perhaps it's actually quite easy to have this type of upbringing, but for several complex reasons the finer points of this are largely irrelevant and more of academic than actionable interest.

What you do develop is a finer appreciation for not egregiously messing things up. And if something explodes-- well, it better have been part of the plan. You better be testing Semtex. Or rocket fuel. Breaking things is part of the agenda, but only when it was part of the plan.

Not so with physics, where breaking things is almost always a mainstay in your remit but almost never planned (there was a big exception on July 16, 1945 in the physics community, and a few times over the decades that followed). Unlike in engineering, this is pretty much always tolerated, and laughed-off. Initially, I chalked this up to some funny quirk of some of my colleagues having the "big physics brain." You're just too clever to care.

Then I came to the conclusion that it is equally possible to suffer from small STEM brain in physics as in engineering, that in many cases, the semantic distinction is just that-- semantic-- and you stand to learn a lot more in many cases if you engineer things than if you physicist them.

On the grand spectrum of Makes Things->Breaks Things, we place physicists moderately close to the "breaks things" extremum. Engineers we situate in the middle, but owing to the fact that actually breaking things (unless they're designed that way) has solidly obtained taboo status in engineering, we place them slightly shy of the 50% mark and shunt them infinitesimally towards makes things. We reserve the rarified air of purely "makes things" for the technicians or machinists who actually have to put the nuts and bolts together and make sure anyone has the parts they need to play around with strange and lofty ideas.

My grandfather got a degree in English from Hartford (note: not Harvard). Then he was a machinist and a milkman. Sometimes I think it's some sort of inter-generational parable calling out to me from the '80s. Sometimes I think it's just one of those quirks of life.

The "IED" On My Kitchen Window (2020)

In Spring 2021, I was nineteen and depressed. I had completed only a single term of university education, which was (necessarily) heavily restricted due to a once-in-a-century global outbreak of respiratory disease and which I had spent mostly in bed, motionless, on ibuprofen, sleeping or making my way through my family's shared Kindle library, which meant that I made my way through all 1000+ pages of Stephen King's The Stand in December without making similar progress in special relativity for my first set of physics exams in January, which I botched in spectacular fashion (they were unofficial and only enforced at Trinity). I sat them remotely, because we were informed that, actually, Lent Term was cancelled for pretty much everyone in the UK that year, and by the time I saw anyone again in April I had forgotten most of what had brought us together in Michaelmas 2020.

This reprieve gave me some crucial time, however, to do a couple of things. We were often reminded, over and over, that, yes, Newton had gone here only a "mere" few hundred years ago and, yes, had similarly been expelled from his rooms due to a pandemic (though in his case it was a bubonic outbreak of slightly more biblical proportion) and that he had managed to cobble together calculus during his exile, so we should look on it as some sort of opportunity, while ignoring the fact that Newton had to pay a lot less for his education than we did, and at least his exams were a stretch easier because he himself had the luxury of not having to revise calculus, only invent it.

One of the things I did was write 60,000 words of a novel manuscript. That's the last time I'll mention it, because only the first 40,000 words were any good, and only c. 30% of those 40,000 words were publishable, and only 10% of that 30% would have actually survived an English supervision at Cambridge.

One of the other things I did was start messing around with a bunch of old electronics. I had been mostly a software person as a kid, but I (like many kids with my background as an isolated teenager) knew how to solder and how to program, and had messed around with Raspberry Pis a little bit when I was small, so it was fairly easy for me to wire up some LEDs and LCD screens and resistors to a breadboard and program them with an Arduino, all of which I brought back to Cambridge with me in order to set up a little display on our floor's gyp window.

Let's backtrack a second, because "gyp" isn't actually a word outside Cambridge, where it means something like "kitchenette," but Cambridge maintains this antiquated synonym because that was the old name for servants that rich students use to have with them during their studies (I know, I know). Also, you need to have some knowledge of the geometry of the Wolfson Building at Trinity, which was an impregnable brick-and-glass eyesore responsible for housing freshers (First Year undergrads) with limited natural lighting, no internal ventilation, and two staircases that ran up and down the building like two brutalist spinal columns, shielded in glass, our rooms and corridors the cramped ganglia which ran off them. The building is two halves, each a near-perfect image of the other. You can hear anyone going up or down the staircase from any room, and you can see everyone from your gyp, which also has little glass panes. Note that none of these pieces of glass provide any view to the outside world; it's all internal. Watching people move up and down, down and up these staircases, all from your gyp window, is the best method you have of learning faces during COVID at Trinity.

Everyone on the staircase can see you in your gyp. COVID rules mandate only one person per gyp per time (a rule obeyed by no one, ever), and everyone needs to use the gyp, so you learn to recognise people by, (a), their presence in their respective gyp and (b) the colour scheme of that floor-- and yes, each floor had a different "colour scheme," but this was limited to the kitschy boards-slash-plaques outside our rooms where our names were painted at the beginning of each year.

C floor was my floor. It wasn't the good floor, like B floor, where there were fun and well-dressed humanities people who were wealthy, and manifested Byronic ideals of youthful beauty. Nor was it like E floor, which was rumoured to have mini-fridges and, crucially, got natural UV exposure from the two arched skylights that served as capstones to our two drab brick stairwells.

But C floor had orange plaques, and it had me. Which meant that it could have its own, fully-automated, 9 V-powered LCD screen, framed by flashing LEDs in primary colours, and which displayed our names and a cheeky greeting that you were, indeed, on C floor, the "best" floor, which was the "domain" of "Raymond Ramm and X and Y and Z" (I went by my middle name back then, and X/Y/Z and I don't really talk anymore, so I won't name-drop) and which survived about 48 hours. This was long enough that someone whom I wanted to impress briefly entertained me as I explained to them, desperately and in agonising detail, while they were looking for someone else, how I built it and how cool it was.

This brief spell of performative engineering was concluded when two porters knocked on my door on the eve of the second day. They asked me if I was the "device's" architect, whether they could take my name down in case they pursued disciplinary action, and whether I was aware that my creation resembled an "improvised explosive device," a charge to which I almost responded with the question of how many explosive devices, in their experience and wisdom, greeted their victims with the full names of its assemblers, before I neglected from doing so out of the realisation that all porters could snap me like a dry twig, and here in front of me were two of them, and so the finer points of debate were best left unexplored.

But engineered indeed the thing was. That was when I finally concluded that, one, Cambridge (or at least Trinity) were not for me and potentially never would be, and, two, it's always cooler to be the one who's getting asked to take down their breadboard than be the person asking what exactly the glowing, wiry thing is. Better to build than criticise.

Glass Tubes That Glow

I spent most of the past year working on a plasma physics experiment in the realm of laboratory astrophysics, despite my original intentions to do theoretical/computational work in the domain of inertial confinement fusion. The long and short of this project is that we made a glass tube containing a state of matter similar to what you'd find in lightning, then put a slab of graphite in front of it. When we fired CERN's proton beam at that target, it made a bunch of "secondaries"-- eventually, electrons and positrons. We wanted to see what happened when these passed through the "lightning in a bottle." I gave everything a good go, lasted a year, and managed to write a paper about the code I developed, and which we've recently submitted for peer review. So not by any means a wasted year.

Big bonuses came in the form of going to CERN four separate times, so much so that some quick math revealed that ~ 20% of my year was spent in and around the Swiss-French border outside Geneva, a location I now know with startling familiarity, though with only a little tenderness, as most of my time here has either been spent working 12+ hour days, or alone for a week or two on end. Obviously, it goes without saying (though I'll say it here) that the opportunity to actually go to CERN, write code there, and spend a lot of time working directly with accelerators, is an amazing privilege that many physicists would be a bit jealous of.

Things didn't work out. My fascination with particle physics was really limited to the theoretical-phenomenology side (for those of you outside of high-energy physics (HEP, i.e. 'particle physics'), particle phenomenology is lightweight theoretical physics; you care about data from CERN, but don't really spend a lot of time worrying about setting things up, is my understanding-- compare to pure, pure theorists, who likely don't care much one way or the other what is happening here, because until HL-LHC, all of this is below the energy scale of interest, anyway). I wasn't concerned much with the ins-and-outs of the experimental side; if you want to do experimental physics on LHC data, the world is your oyster. Pretty much anyone with a decent enough physics background can, if they want, probably find a PhD position somewhere with ATLAS in the title-- that's how massive and far-reaching the collaboration is.

The devil is in the details. It's not all what it's like in the movies. A lot of students who "work" on experiments at CERN don't actually spend much, if any, of their time there, especially during long shutdowns (like the one that is about to happen), and the work you're doing is not romantic. You're a data rat. A smart data rat (you need to have a good knowledge of the standard model, Python/C++, and math in order to make sense of what you're seeing), but still a rat. Most PhDs are somewhat like this. If you're smart and lucky enough, you can be a theoretical physicist, but most of these positions are closed to people who didn't do mathematics or mathematical physics/theoretical physics for their MSc, so this door (mostly) has been closed to us since day 1 of our physics education.

There are so many people who like being a rat. Most STEM PhDs involve some form of rat work, either in the lab, or at the computer. That's why it's tough being a PhD student. The hours are long, the work is really complicated (much harder than a lot of the work you'd get at a far better-paying gig), and the life is really isolated. Despite this, there are some (many) of us out there who actually couldn't survive in the real world without this constant frustration, friction, obscurity, and novelty. You might earn more money working in the real world. But you don't get to work as a physicist at CERN in your 20s.

So the premium you pay in money is made up for in some interesting life experience. And I don't think that's a small thing. If someone asks you in your 40s what you were up to at 23, it's probably more awe-inspiring, if you're a STEM person, to talk about a PhD in particle physics than quite a few of the alternative avenues (Excel spreadsheets, banking, etc.). To people who didn't go down that route, they may think that doing a PhD would have been intermittently miserable, and that they're much happier with a house and an earlier start to their career. And they might, in fact, be right.

Which brings me to the concluding point: making money, just like a PhD, isn't for everyone. Most of us in academia only need/care for enough money to keep the water running and (occasionally) let us upgrade MacBook, but other than that, I think most of us would be hard-pressed to spend money. If most of us doing PhDs were handed one million dollars, I imagine most of us would carry on doing PhDs, albeit maybe in nicer clothes and in bigger apartments. Though even then, on the upgraded living conditions side, I'm not sure.

"Doctor"

It's objectively ridiculous that you call yourself "Dr.", in theory, after getting a PhD. Personally, I don't have any plans to advertise my PhD anywhere other than on my CV (I think I'd die if someone called me anything like "Dr. Ramm", eww), and after your first postdoc or so, it no longer really counts for much, anyway. By then, if you haven't accrued a respectable publication record demonstrating genuinely-novel research, you're unlikely to be able to fall back on institutional prestige or anything so superficial.

Problem is that a PhD is basically a requirement now for most cutting-edge ML jobs. And, increasingly, in STEM, it's a desired qualification in other sectors (banking, software engineering), which is a bit weird, because until the fairly recent past, a PhD wasn't a "vocational" qualification unless you had plans for entering academia. But it makes sense: we don't really teach "ML" at undergraduate level as such, at least not to the same level we teach chemistry or physics (and these are often the people who will benefit the most from ML). And, after getting a master's in these subjects, most of us are still years away from knowing enough to really work in ML in any serious capacity, other than as data scientists. E.g. Bayesian inference, which is essential bread-and-butter stuff in ML but which you only encounter very briefly in physics, and you're too busy doing problem sets to be building neural networks, unless you're lucky enough to bag an ML project for your master's².

Most people working in hardcore ML have PhDs in CS or Engineering. The Machine Learning Group at Cambridge only even hands out degrees in Engineering-- not CS or statistics or anything else-- and most people at Oxford in OATML have some explicit engineering background, even though their domain is registered under the Department of Computer Science. The component of your PhD dedicated to ML needs to be significant, not minimal, if you want to be a competitive applicant in the postgrad market, and those types of projects are heavily concentrated around engineering and CS right now, rather than in physics where, if they are used, it's results-oriented and not overly concerned with building new types of neural net, or anything like that.

fin

Making things versus breaking things is a big motivator for people in engineering; coding neural nets provides a very malleable, novel, and powerful space for doing this from your desktop. And, because your workload has been shunted off into the space of software, not hardware, there's no real incentive to "break" anything in the physicist sense, though there are of course benefits to probing the limits of what you're doing, and locate failure cases. That's not really "breaking" things in the traditional sense. It's more "pre-breaking" them, or breaking them in God mode where you know what's happening and can easily fix it before you hand it off to a person/people who didn't work on the project with you and likely have no idea what the hell you've actually been doing.

I guess I just got saddened by breaking things all the time. Or watching people shrug off failure cases, or glacial code, or freakish behaviour in our bottle of plasmas. You actually waste a lot of time when you don't slow down to try and understand why things are breaking. And, best case scenario, you'll have to relive the whole experience again when you return the next year, or for the next experiment. And you don't actually learn anything either-- except the (wrong) lesson that bodges are better than fixes.

Never trust a physicist. Whatever the overtrained, overconfident equivalent of "butter-fingers" is, they have it, some cosmic and God-granted capacity to destroy and disassemble-- the living embodiment of entropy, and the universe's inexorable march towards heat death. Physicists are only good for speeding that process up.