Tarrosion

SecureBio Detection | Senior infrastructure engineer; [senior] software engineer (high-performance pipelines) Cambridge, MA, USA (metro Boston) | ONSITE strongly preferred; REMOTE (US east coast) possible for exceptional candidates

SecureBio is a nonprofit working to protect the world from catastrophic pandemics. We operate the world’s largest metagenomic biosurveillance network, which performs daily sequencing of wastewater samples and nasal swabs to screen for emerging pathogens. Our computational team develops detection pipelines responsible for rapid processing of deep sequencing data in order to flag potential threats.

I'm the hiring manager for two roles:

- Senior infrastructure engineer, to own cloud infra, devops, CI/CD, AWS security, etc. https://securebio.org/careers/2026-cloud-infra-engineer/

- [Senior] software engineer focused on high performance data pipelines. https://securebio.org/careers/2026-software-engineer-pipelin...

Prior experience with biology or bioinformatics is not required. If you're interested, please apply via the relevant job posting. Feel free to drop me a line with questions; email is in my profile.

https://securebio.org/careers/

SecureBio Detection | Senior infrastructure engineer; [senior] software engineer (high-performance pipelines)

Cambridge, MA, USA (metro Boston) | ONSITE strongly preferred; REMOTE (US east coast) possible for exceptional candidates

SecureBio is a nonprofit working to protect the world from catastrophic pandemics. We operate the world’s largest metagenomic biosurveillance network, which performs daily sequencing of wastewater samples and nasal swabs to screen for emerging pathogens. Our computational team develops detection pipelines responsible for rapid processing of deep sequencing data in order to flag potential threats.

I'm the hiring manager for two roles:

- Senior infrastructure engineer, to own cloud infra, devops, CI/CD, AWS security, etc. https://securebio.org/careers/2026-cloud-infra-engineer/

- [Senior] software engineer focused on high performance data pipelines. https://securebio.org/careers/2026-software-engineer-pipelin...

Prior experience with biology or bioinformatics is not required. If you're interested, please apply via the relevant job posting. Feel free to drop me a line with questions; email is in my profile.

https://securebio.org/careers/

I'm curious to see how Claude can interact with Blender, and how people use it. I use Claude every day for both work and personal research, overall think it's a great product, but I've found it (thus far, never bet against generation n+1) remarkably terrible at spatial reasoning. That seems pretty key for Blender!

This author is really prolific in the Julia visualization ecosystem. One of my favorite projects from them is actually quite useful outside Julia: a super minimal but slick Unicode character lookup, glyphy.info.

[citation needed] that some combination of "New Urbanism, traditional neighbourhood design, streetcar suburbs, one-way streets, bike paths, walking paths, mixed-zone walkable villages (light commercial with residential), smaller single-family houses and duplexes, triplexes, houses behind houses." is not in fact optimal! (For certain objective functions)

sparsely updated blog: https://evanfields.net

Is multi-agent collaboration actually useful or am I just solving my own niche problem?

I often write with Claude, and at work we have Gemini code reviews on GitHub; definitely these two catch different things. I'd be excited to have them working together in parallel in a nice interface.

If our ops team gives this a thumbs-up security wise I'll be excited to try it out when back at work.

I'm finding 4 Opus good, but 4 Sonnet a bit underwhelming: https://evanfields.net/Claude-4/

I can confirm this has been true since at least v0.4!

I'm sorry for your loss, and I hope that helping others through this project helps you find some solace. IMHO, it's a mark of character that your response to having a problem is "I want to help other people so they suffer this problem less than I did."

Whoops forgot the link https://entropicthoughts.com/the-misunderstood-kelly-criteri...

This makes sense in the context of trying to maximize log wealth (or I think any concave function of wealth, though the arithmetic is different). But in one of OP's other articles [1], he says the Kelly criterion doesn't require trying to maximize log-wealth, that this is just a common misconception -- all that's required is maximizing something growing geometrically over time.

This I don't understand, maybe someone help me out? Say the real growth rate of capital (or interest rate available to me, whatever) is 2%/year and I have a 10 year time horizon. So $1.00 today is ~$1.22 in 10 years. More generally, if I have wealth X today I will have 1.22X in 10 years. And if X is not a constant but a random variable and I want to maximize future expected wealth (not log wealth), that's just max(E[1.22X]) and by linearity of expectation I should just maximize wealth today to maximize in 10 years time.

So Kelly being appropriate must have some other conditions, right? Wanting to maximize log wealth is surely sufficient (and individually probably ~rational). What else?

[1]

OP, you have a comment there about mouse support in the Keyboardio. I've been using a Keyboardio since 2019 and haven't much tried the mouse support -- any advice? How did you set it up?

I'm curious what kind of slow IO is a pain point for you -- I was surprised to read this comment because I normally think of Julia IO being pretty fast. I don't doubt there are cases where the Julia experience is slower than in other languages, I'm just curious what you're encountering since my experience is the opposite.

Tiny example (which blends Julia-the-language and Julia-the-ecosystem, for better and worse): I just timed reading the most recent CSV I generated in real life, a relatively small 14k rows x 19 columns. 10ms in Julia+CSV+DataFrames, 37ms in Python+Pandas...ie much faster in Julia but also not a pain point either way.

Why don't gradients vanish in large scale LLMs?

How do modern foundation models avoid multi-layer perceptron scaling issues? Don't they have big feed-forward components in addition to the transformers?

Location: Cambridge, MA, USA [Boston/Somerville in case anyone is text-searching]

Remote: strongly prefer in person or hybrid

Willing to relocate: I'm very happy in the Boston area, so only under extraordinary circumstances

Technologies: data science, machine learning, technical communication, simulation, optimization, physically motivated models, heuristics; Julia, Python (Django, numpy, scipy, sklearn, cupy, pandas), AWS, Postgres/Postgis, Snowflake, Gurobi, NLopt, etc.

Resume: https://www.dropbox.com/scl/fi/wevhopcf81vx6vdoarn5a/efields...

Email: in the resume

Blurb: PhD in operations research -> startup employee #2 -> scale startup, gradually moving from data science IC to VP of data science + software engineering -> now looking for mission-meaningful hands-on technical work. I'm particularly good at technical communication as well as translating from the physical world (science/business) to math models. Open to companies of any size, especially motivated by the biosecurity, biotech, robotics, and clean energy sectors.

Something cool I built that's not work related: https://nomai-writing.com/ in the spirit of Outer Wilds

Location: Cambridge, MA, USA [Boston/Somerville in case anyone is text-searching]

Remote: strongly prefer in person or hybrid

Willing to relocate: I'm very happy in the Boston area, so only under extraordinary circumstances

Technologies: data science, machine learning, technical communication, simulation, optimization, physically motivated models, heuristics; Julia, Python (Django, numpy, scipy, sklearn, cupy, pandas), AWS, Postgres/Postgis, Snowflake, Gurobi, NLopt, etc.

Resume: https://www.dropbox.com/scl/fi/wevhopcf81vx6vdoarn5a/efields...

Email: in the resume

Blurb: PhD in operations research -> startup employee #2 -> scale startup, gradually moving from data science IC to VP of data science + software engineering -> now looking for meaningful hands-on technical work at a company inventing something in the world of atoms

Something cool I built that's not work related: https://nomai-writing.com/ in the spirit of Outer Wilds

I don't think I follow, can you elaborate?

* Car users are quite dependent on the government for transportation, e.g. the many billions (in the US) of public dollars spent on roads each year. Hard to get around without them!

* Even if we hand-wave that away and assume car users aren't dependent on the government for transportation, surely _everyone_ is dependent for other reasons like enjoying public goods (national defense, clean air and water, rule of law), access to the social safety net, etc.?

I've heard this before -- that oversized cooling units (whether standalone AC or part of a heat pump) mean muggy interiors in the humid seasons. But...why? I'd think that a fixed amount of air compressed in the compressor means a fixed amount of condensation runoff from the unit, and it wouldn't matter much whether it's a big unit running occasionally or a small unit running frequently. Why is that wrong?

Anyone have suggestions for the lowest friction / best UX way to generate and study Anki cards? (my smartphone is Android, if that matters.)

I know spaced repetition is super helpful and I should be making and study cards to help with language learning and other topics I'm studying, but it always feels like a slog to try to find a deck (which won't end up being what you want) or manually make a bunch of cards, the UI is a little meh, etc.

What's the user-facing difference between Lux and Flux?

I agree that cars and trucks are appropriate for a variety of trips and cargo-hauling use cases that scooters are not appropriate for. But I confess I'm missing the implication - what does this imply about the claim that micromobility is held to a standard we don't typically hold other transportation modes to, in part because we have a default cars-first perspective?

Bird the company has/had lots of problems, but I'm sad to see them hit bankruptcy for what it signals about micromobility as a whole.

I've long been frustrated that micromobility is held to a standard no other transportation mode is expected to meet, at least in the US. E.g.

- "Micromobility doesn't make any sense, you can tell because all the companies lose money." But every other form of transportation is deeply subsidized! In the US we spend hundreds of billions of dollars per year on roads, trains get federal infrastructure money, public transit [where it exists] isn't close to fully paid for by fare revenue, airlines get bailed out in recessions, all carbon-burning transport has an unpriced carbon externality, etc.

- "I hate the scooters, they clutter sidewalks." Often the clutter is real and a problem and I wish people behaved better. Among the people who I wish behaved better, I'd include car drivers who park in bike lanes or on sidewalks, yet the prevalence of such behavior doesn't create widespread calls to ban all cars. (Well, in some corners of the internet...)

More generally, it's weird that scooter companies are liable - legally or de facto - for such behavior at all. If I get a rental car from Hertz or similar and then park illegally, nobody claims this means Hertz should get a ticket or be shut down. Why are scooters different?

At the end of the day, in the US we've built our legal regime and land use patterns around the assumption that most travel should be by car and this travel should be subsidized. The fact that scooters create some frictions by taking up some of the space left over by cars is more an indictment of the system than of scooters.

Am I missing something, or are all these effect sizes just shockingly improbably large? "decreased the risk of all-cause mortality, CVD-related mortality, and cancer-related mortality by 23, 26 and 22 % "

1/4 drop in mortality just from fiber seems...like too much? Am I missing something subtle about how these mortality reduction effects are defined? Or perhaps these are conditional expectations but not causal effects? (Example: if eating fiber is negatively correlated with binge drinking alcohol, then P(death|fiber) probably << P(death|~fiber), but it's unclear a priori how much of that difference is due to alcohol)

http://evanfields.net

Sporadically updated, but I'm pleased with most of the posts. It's a mix of personal reflections and some personal tech projects.

Mostly I've been a huge fan and think I learn a lot, but occasionally his explanation doesn't click with me. One example is his Fourier series where I want to think about a change of basis and multiplying an input signal by "reference" functions, but his explanation is more about circles spinning on circles.

Yeah, I find it super useful when you want a maximally quick look at the shape of your data. Even if I'm local (ie not SSH'd onto a remote system where displaying graphics could be a hassle), using UnicodePlots can "break my flow" less than generating prettier plots in another window.

The discussion around shared e-scooters (and similar micromobility) tends to have a ton of hidden status quo bias. For example, one of the most common complaints about e-scooters is that people leave them all over the place and they clutter sidewalks. This can be a true and valid complaint!

But people do that, in part, because cities don't provide good alternatives. A city the size of Paris probably has literally millions of on-street parking spots, or at least hundreds of thousands. (E.g. NYC has ~3 million [1]). Each car-sized parking spot can fit 12 scooters. So replacing 5% of parking spots with micromobility parking would provide space for something like 500,000 shared micromobility vehicles. For context, that's order of magnitude larger than the number of e-scooters in Paris.

(FWIW, Paris is maybe not a perfect example here; it's a relative leader in reclaiming space from cars for shared mobility and biking and such. Also when I was there last summer the scooters weren't a nuisance at all because they were overwhelmingly parked in marked areas, not littering sidewalks. YMMV.)

But especially in the US, we're so entrenched in thinking that only cars can be real mobility that we can't imagine shifting even 1% of the space and money we spend on cars towards more shared and less polluting forms of transportation. It sucks.

[1] https://gothamist.com/news/how-else-could-nyc-use-its-12-cen...

Do you worry about ebike theft? If not, how do you lock it in public?