Joseph Lyon

Hello LinkedIn friends, I’m still on the market. Trying to find a new gig. For personal reasons I need to remain remote, but the occasional trip into NYC or an offsite meeting is fine. I’m looking for IC roles at this point. I want to build cool shit that helps people. I’m great with TypeScript and Node.js, as well as data engineering. Full stack or backend is fine. If you know of any roles I’d greatly appreciate a recommendation or even a link sent via DM. If you’re a resume wiz, maybe you could review mine to see if it needs work. The tech world is all about ghosting candidates or sending out broken (([candidate.name])) rejections. I appreciate any support! #opentowork

24

4 Comments

Carta's data is ....off? Coastal Recruiting communicates with hundreds of engineers weekly, primarily in New York and San Francisco. We consistently encounter software engineers with 2-6 years of experience requesting base salaries between $180,000 and $250,000, plus equity. Perhaps "misleading" is a better term, as I frequently hear founders saying, "But Carta indicates we can hire this engineer for $140,000 to $160,000." We then observe Seed/Series A companies paying $200,000 to $300,000 for top talent. The disconnect may stem from the fact that companies raising funds from Sequoia, Khosla, Founders Fund, Lightspeed, a16z, KP, Greylock, Index, NEA, Lux, and others are not seeking what 95% of the companies providing data to Carta are seeking. These companies are looking for the top 0.01% of engineers. I say all this while also acknowledging my appreciation for Carta. I value the data and charts it provides, but it's important to remember: "There are three kinds of lies: lies, damned lies, and statistics."

26

9 Comments

I think we are going to overhire AI engineers leading to massive amount of jobs but also because we are not clear where the results are. I see two kinds of engineering workforces being deployed to figure out these unknown outcomes: 1. Training and Scaling We are approaching limits of current scaling methods and the next set of outsized gains requires a step function improvement in data center deployment sizes which wont come easy. We will spend large amount of engineering resources in variations of training LLMs. These could involve trying new variations on RLHF, multi-modal, mixture of experts etc. A lot of territory in current transformers hasn't be expanded wide enough and we are yet to find out if there are any gains in these offshoots with current hardware. 2. LLM Meta programming We call these by a variety of popular terms now such as - agents, orchestrations, RAG etc but they all primarily are paradigms of LLM meta programming. There are way too many tools and frameworks for these and in the short term default scenario we will be over-engineering and increasing complexity before we know how to cut and simplify.

11

Well said. The alarmism around AI replacing software engineers often overlooks the complexity of our role. While AI tools like Devin are impressive, they are assistants, not replacements. 𝟏. 𝐂𝐨𝐝𝐢𝐧𝐠 𝐢𝐬 𝐣𝐮𝐬𝐭 𝐨𝐧𝐞 𝐚𝐬𝐩𝐞𝐜𝐭 𝐨𝐟 𝐬𝐨𝐟𝐭𝐰𝐚𝐫𝐞 𝐞𝐧𝐠𝐢𝐧𝐞𝐞𝐫𝐢𝐧𝐠. 𝟐. 𝐒𝐞𝐧𝐢𝐨𝐫 𝐫𝐨𝐥𝐞𝐬 𝐢𝐧𝐯𝐨𝐥𝐯𝐞 𝐚𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞, 𝐝𝐞𝐜𝐢𝐬𝐢𝐨𝐧-𝐦𝐚𝐤𝐢𝐧𝐠, 𝐚𝐧𝐝 𝐜𝐫𝐨𝐬𝐬-𝐭𝐞𝐚𝐦 𝐜𝐨𝐥𝐥𝐚𝐛𝐨𝐫𝐚𝐭𝐢𝐨𝐧. 𝟑. 𝐔𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝𝐢𝐧𝐠 𝐛𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐜𝐨𝐧𝐭𝐞𝐱𝐭 𝐚𝐧𝐝 𝐝𝐞𝐟𝐞𝐧𝐝𝐢𝐧𝐠 𝐭𝐞𝐜𝐡𝐧𝐢𝐜𝐚𝐥 𝐝𝐞𝐜𝐢𝐬𝐢𝐨𝐧𝐬 𝐚𝐫𝐞 𝐜𝐫𝐮𝐜𝐢𝐚𝐥 𝐡𝐮𝐦𝐚𝐧 𝐬𝐤𝐢𝐥𝐥𝐬. 𝟒. 𝐄𝐱𝐢𝐬𝐭𝐢𝐧𝐠 𝐜𝐨𝐝𝐞𝐛𝐚𝐬𝐞𝐬 𝐚𝐧𝐝 𝐫𝐞𝐯𝐞𝐧𝐮𝐞 𝐬𝐭𝐫𝐞𝐚𝐦𝐬 𝐫𝐞𝐪𝐮𝐢𝐫𝐞 𝐜𝐚𝐫𝐞𝐟𝐮𝐥, 𝐡𝐮𝐦𝐚𝐧-𝐠𝐮𝐢𝐝𝐞𝐝 𝐞𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧. AI will definitely reshape our workflows, but it won't replace the need for skilled engineers. Instead, it may allow us to focus more on high-level problem-solving and innovation. Arvind Telharkar 👏

7

1 Comment

More ramblings on #techinterviews. Let's talk about homework assignments and how they aren't a favorite for interviewees. The times I've avoided them (as in didn't continue interviewing with the company) they were in addition to multiple rounds of interviews. They also weren't small in effort. So the expectation was around 3-4 hours of normal interviews along with at least that much time doing a homework assignment. Six to eight hours is a big ask of your interviewee. Granted, if you fly them out for in-person interviews, 4-6 hours might not be unreasonable (but you better feed them lunch.) I can recall three that I actually did and the approach made it mostly worthwhile. All three had some aspect of the problem that directly tied to the day-to-day work you would be doing. One turned out to be about 25% of typical day-to-day work, including the library that they used. All three had a person-to-person interview going over your code to see what you were thinking/why you did something a particular way, etc. In the 2 out of 3 that I continued with, it was very collaborative and more like a code review of a peer than trying to show you up. I dropped out of one interview for other reasons (probably a topic for another day.) Those experiences actually moved me from mostly despising homework interviews to being open to hearing what the process is and if I'm ever in the position again to define interviews, I might consider give one. I think they could be a good compromise to finding out the coding and problem solving abilities of a candidate without resorting to leet code exercises. Make it short and related to the types of problems they will be working on. Added bonus if you have a starting point for them to work with. Very few projects are 100% greenfield so seeing how they interact with "legacy" code is a good thing to do. Have them set up when the follow up will be. If they need a month for something that is 2ish hours, something is fishy. Since things can come up, you might end up with the opportunity to see how they communicate missed deadlines. I actually had to do that with one of mine. Work got busy and I didn't have as much free time as I thought. I rescheduled early as to give them time to adapt. They appreciated me being upfront. Pretty much a free soft-skill interview on both sides. So in summary, homework questions can be good but they better be about 1/4 to 1/2 of the overall interview in length (with the total interview time around 4 hours.) Make it relevant to the job they will be doing. Make it easy for them to run locally.

9

1 Comment

I’ve been doing a lot more recruiting and one trend I’ve noticed is a tendency to talk about using a framework when really describing a process. There are two key differences: Frameworks are structured but open ended and are not prescriptive. Processes are prescriptive sequential steps. It’s important to identify when you want to use one or the other: - Coaching an engineer: Framework - Planning: Probably framework with a tiny bit of process for scheduling - Recovery of a data center: Process A more engineering centric / systems view would be to look at frameworks as the interfaces in engineering and process as the implementation behind the interface. Too often we’re focused on the implementation of the process before we’ve established what the framework is we want to use.

50

3 Comments

> Leaders forced to rely on outdated metrics like lines of code to gauge productivity. "lines of code" is still a thing?? I should find one of those organizations, I can write hundreds of lines of junk in an hour. Just promise I'll never have to maintain it. 😜 Seriously, software development management and developer productivity measurement are hard problems. Over thirty years and I've yet to see a simple numeric metric that didn't cause more harm than good. A good developer recognizes the engineering tradeoffs and matches them to the business' requirements for each task. No simple metric captures the complexities involved.

3

🚨 Check it out - hot off the press! 📰 Sudeep Das, Pradeep Varma and Ying Yang are all hiring as we continue to scale! 🙏 If you have a passion for Machine Learning and expertise within Deep Learning, Computer Vision and/or NLP domains, let's connect! #experimentation #DoorDash #hiring #machinelearning

9

“There’s no jobs for Jr Software Engineers out there!” NOT TRUE…But there are some major issues in today’s hiring market. 1. MISALIGNED FOUNDATIONAL EXPERIENCE: The majority of aspiring SEs are either: Recent CS grads (less hands-on exp); OR, Bootcamp grads (less theory). Meanwhile...Most enterprise companies who hire a ton of SEs are looking for CS grads that ALSO have hands-on experience already. 2. MISALIGNED TECH STACKS: The majority of aspiring SEs are only experienced in NodeJS and/or Python. Meanwhile…Most of the companies who hire a ton of SEs are using Java or C# backends. Companies using NodeJS or Python backends are often smaller and can't hire as many SEs. That’s why our Enterprise Accelerator is making a difference. We identify diverse CS grads who are hungry, and we teach them the hands-on Java skills that companies are using in the real world over 3 months. Best of both worlds 😎 If you're interested in seeing some of this kind of talent--send me a DM and we'll get you connected. Want to join our next Enterprise Accelerator cohort? Stay tuned for the next webinar in early August.

77

7 Comments

I try to avoid political posts on LinkedIn, because there are people I work with who I genuinely enjoy working with.  They're smart, great at what we do, the type of engineer you want on your team, and most likely we strongly disagree on political issues that are important to both of us.  If we started discussing politics we could destroy a great professional relationship.  I want to be able to network with those people on LinkedIn without them needing to know the details of my political beliefs just like I have no interest in knowing the details of theirs.   Personally the only time I ever disconnect with people on LinkedIn is when they make political posts I strongly disagree with, because I don't want to see that on my LinkedIn.  If I wanted to see political posts I disagree with I'd go on Facebook.  

33

31 Comments

My thoughts on bypassing dumb biases when trying to get your next tech position... The "overqualified" bias: If you are having trouble finding your next job because folks keep saying that you are overqualified, trim some stuff off of your resume. Seriously, take 5 years or 10 years off. You do not, in fact, have to give your whole work history. If you keep getting the message over and over that you are over-educated, trim back on that. Your next tech position does NOT need to know that you have a doctorate in music or french literature. If it isn't helping - strike it. The "they'll say anything to get a new job because they were laid off and aren't working" bias: I had this happen to me one of the first times I was laid off. I had someone (chiropractor) very interested in hiring me. The interview had gone very well and on the way out the door he asked what my schedule was with my current employer because he wanted to schedule a follow up (I was doing contract massage therapy at the time). So I told him that my current gig had had to cut back on the therapists they had due to less clients scheduling sessions. It had been maybe 3 weeks since I had had a session and I had been cut because I was one of the newer folks. "OH, I see." That was literally the end of that. The interest fizzled up and died, right there. So... do contract work. The first time I was laid off from a tech job I had a friend that needed a web app set up for their business to help advertise a salon but couldn't afford regular rates. So we traded - I gave them a discounted rate and traded for some of the businesses services... yep got my hair done several times. I later found a similar contract with a friend trying to expand their existing website offerings and get better results in google. This was without the trade but it was helping a smaller business at a reduced rate. The point being that as I started getting serious interviews some months later, I had no gaps in my employment. No, I hadn't made good pay at either of those "jobs" but I filled in the resume gaps while my "day job" was mostly trying to get new employment. Not enough Experience bias: A similar thing can be done for folks who don't have enough experience. If you are working in something else but trying to get a job in technology/IT, moonlight or trade with a friend or local business at a reduced rate. You'd be surprised at how much value you can get out of a trade. The only caveat being, make sure you do your best to honor the agreement. Don't treat it as a throw-away contract -which is also why you can/should be picky when offering services at a reduced rate. You will very likely want this person to be a reference for you when you go to get your next job. Make that clear and only trade with someone that you believe will keep up their end of the bargain. If you have family with their own business - seriously... go do website work or IT support for them and list that. Good luck out there.

4

For my eng managers / rusty engineers out there, tell me if this sounds familiar. After some meetings get rescheduled, you suddenly realize you now have a two hour block of free time(!). You decide to try and fire up your dev environment to make some small changes or dig into a new service you’ve been eyeing for a while. As you start you environment, you immediately run into configuration issues and errors with an upstream service. After asking around and tinkering for a while, you manage to get your services up and running. You try to make some changes and want to test them out, but someone else is using staging and you have no idea if you’re PR is going to break a downstream dependency. You try to manually route where requests would go in the codebase and analyze unfamiliar unit tests. You look at the clock and realize it’s time to get back to other meetings, leaving your PR for yet another day. So many engineers run into these issues on a daily basis, leading to teams relying only on unit tests for changes or abandoning something that should have been a quick fix in an unfamiliar area. At Crafting, we build our sandboxes so they seamlessly integrate with all your services, allowing engineers to get up and running instantly before testing against a shared staging environment without having to worry about manual configurations and coordination. Stay tuned for more exciting news coming soon!

72

9 Comments

A lot of people turn down offers due to compensation. That’s a fact. I didn’t make that up, it’s not a personal opinion, it’s an important part in making a decision. People want more transparency around compensation, so here is what I’ve been seeing on my end. I’ve been recruiting software engineers with 4+ years of industry experience. On average I am hearing people say they are targeting $180-220k total compensation. I have had a few outliers outside of that. I have people who request $120k and $150k. I’m not sure if those numbers are publicly listed somewhere, but on the lower end those are the two I get. 9 times out of 10, candidates will say they are looking for “the market rate”, “willing to negotiate”, “want to wait til the offer to discuss numbers”, and they will try to counter the total comp expectation question back to me. When we talk some more - they open up that they are willing to take less, that’s their IDEAL* compensation, and I always jot down the comp goal / never ask what they’re currently making. I believe current salary is irrelevant because you are taking a risk when joining a new organization, leaving your tenure, and should look for increase in your next role. I do have candidates who are my OUTLIERS - they request $300k+ with the max so far being $440k. In these situations, I used to let candidates know we probably wouldn’t be able to come close to that or compete. Lately though, I’ve been questioning that logic. I think Microsoft CAN compete, but you have to look at your package holistically. In the last year I have challenged myself personally to see… can I make more investing than my base pay? Can I beat out these Engineering salary expectations without being in the field/ making less? And if I can do it … can I share how* I’m approaching this because yes …. Base pay I don’t make 300k. No where close to that. But despite the ups and downs of the market, I’m determined to clear $300k this year. I believe I’m worth every penny and what those engineers who demand crazy numbers have taught me is …. you’re worth whatever number you want. You just need to get a little creative 🤑 Microsoft sets me up perfectly to do a few things: Receive: 50% 401k match 10% discount through ESPP HSA contribution The numbers below do not account for RSUs granted, bonus pay, Perks+ stipend, or healthcare benefits. My fidelity acct only includes 12 months of investing while at Microsoft, this doesn’t include any money outside of having my current 9-5 job. (This isn’t total savings, doesn’t include any side income, just an experiment.. is this possible?) Since I’m heavy on Microsoft stock/ refuse to sell, I am investing in 4 different categories: 1. Our competitors that pay dividends 2. Safe companies that pay dividends 3. ETFS 4. Outside of tech stocks - I like semi conductors, healthcare, defense and growth stocks

405

49 Comments

I recently saw a "Just write code" variant of the bell curve meme making the rounds and started wondering if there's a better metric for predicting a team's success. There's a very clear signal from my work experience: "Don't read code" - it's also the only development task intrinsically more painful than writing a line of code. As a metric, time spent reading code indicates when to evolve your development process. A prototype with no comments or tests is fine - if you plan to throw it away. To avoid future you reading code, prefer good comments and concise tests to explain and verify what that code is doing. At the system level, writing requirements to describe the next iteration and a design to prove that it can get there are easier to review than the code itself. For my team that got the farthest the fastest with the fewest regressions, I taped system design UML diagrams to my wall and used them as talking points when answering questions. By the second week, the entire team had done the same and were soon creating their own. Onboarding became a no-op as everybody was communicating the same stories with the same pictures. Like all good memes, there is a truth in just write code, but it's never complete.

3

1 Comment

An observation of today's era of builders at the frontlines. I've heard folks talk about the future being in individuals and small teams building products and companies (the creator economy). The market could become more competitive with a flurry of well built products. The reason ppl say is the tools available (specifically AI tools). Anybody can *create* an app today but more importantly, anybody can create a pretty good app. The other reason I speculate is companies are doing much more with less people and employees are very observant. In tech, if sales increases enterprise value by 10mil, they get 1-2mil of it. Eat what you kill. For product/engineers, if they are largely responsible for an increase EV by 10 or 100mil, they still get...their base salary, maybe a bit bigger bonus, juice their equity a tad and hopefully some recognition. This is the standard operating procedure in the industry. It will be curious to watch how employee builder's pay structure changes over the next 5 years.

11

2 Comments

It’s your first SWE job. 11 people are on this video call. And you’re terrified to come off mute. The self-doubt creeps in: 1. What if they realize idk how this API works? 2. What if my opinion is wrong? 3. I bet they’ve already had this idea anyway. But here’s what I wish I’d known: 1. That API isn’t documented, but it should be. 2. Being “wrong” is expected (and encouraged). 3. No one hates hearing a good idea twice. Staying quiet out of fear? Could trap you at Junior for years. Giving others a chance to help you learn? That’s where all the growth happens. —— P.S. Callie and I wanna help you conquer this self-doubt for good. In our free community for entry-level devs: https://lnkd.in/eYrbhdUS

517

50 Comments

"I don't like leetcode style interviews!" See I wasn't making it up! Kurtis Nusbaum tells us more about why engineers don't like leetcode and the types of interviews that lead to a more enjoyable and effective interviewing experience.

4

1 Comment

Hi all, I’m starting a new position on Monday (details coming later), and so my work on olfactory understanding will be taking a back seat to my exciting new position. Smellotron.com will stay up, but I will no longer be actively maintaining it. I’m finishing up a few things before I start: mainly, I hope to get the full version for “Olfactory Label Prediction on Aroma-chemical Pairs” submitted to journals. In between interviews and job applications, I gathered a large dataset of labeled fragrance/flavorant recipes. This dataset will hopefully be made public soon, but in the meantime I put together a demo of generating simple perfume-like recipes on the fly, using a sequence to sequence molecule recommendation model. The underlying label prediction and molecular embedding model is the Open Principal Odor Map model. Frontend design is not my strong suit, and the timeline for cleaning the data and training the model (it’s not even a transformer!) was a bit rushed. Still, I think there is nothing like this publicly available, so maybe you all will find it cool. Full source available at: https://lnkd.in/e4JQhwgj I ended up being friends with a lot of DJs irl, and I intended the UI to capture a bit the magic of using a mixer live. Depending on how busy I'll be in the next few months, you might see (or smell!) me demoing this in a yoga studio or dimly-lit club in NYC. Below is a video of the demo in action, with my collection of aroma-chemicals from the “Beginner's Aromachemical Kit” by Perfumers Apprentice. 

19

This may be the best advice for engineers looking to advance in their career. Product impact will take you much farther than technical mastery. Having a product mindset can easily make you stand out and shine ✨ —and of course it’s great for your business. Technical skills clearly matter for engineers, but being savvy about how to create product impact puts you on a totally different level.

26

1 Comment