Marcin Pawlowski

Want a preview of what I'm going to be talking about in my keynote for RubyConf? You can catch a glimpse of it here!

10

The difference between a decision and a data-driven decision is like the difference between a shot in the dark and a carefully aimed arrow. Which decision do you want to make?

2

Project update! My application tracking extension is now doing fuzzier matching and injecting new information into LinkedIn (and other sites) to keep me from doubling up on applications. Next up: local storage and bulk import/export operations...

12

2 Comments

At InVision, we worked with some amazing software vendors. I know I've talked about most of these in the past; but, I just want to give a final shout-out to some of my favorites: * LaunchDarkly - feature flags completely changed the way we approached product development. And, allowed us to all but eliminate our incidents and outages (shameless plug for my own book: https://lnkd.in/ehxiy6bY). * Postmark - delivered all of our outbound emails and powered our inbound reply-by-email workflows with flawless uptime and befuddling performance. * Pusher - powered our real-time collaboration via WebSockets, bringing our UIs to life and keeping browsers synchronized. * Intergral | FusionReactor APM - gave us deep insights into our JVM performance, allowing us to track down SQL, API, and processing bottlenecks. * Datadog - truly world-class metrics, alerting, and monitoring. * Adobe ColdFusion / Lucee CFML - the programming language and application server that allowed us to move fast, keep things simple, respond to users, and build wonderful things. * Cloudflare - your CDN, WAF, and web workers took a beating so that our origin servers could stay online. * Amplitude - connected me to the users by giving us insight into what our users were actually doing (or not doing). At one point, I think we had over 300 different SaaS subscriptions; so this list is faaaaar from exhaustive. And, I'm sure I'm forgetting other beloved services. And, many of our teams worked with entirely different technologies. This is just my little corner of the organization. And, I couldn't have done it without them. Thank you so much! ❤️

113

18 Comments

Today, we’re publishing our biggest piece yet.   We partnered with Puzzle to write about the next generation of accounting tools.   Our CEOs, Alex Lee and Sasha Orloff, explain how AI can turn accounting from a repetitive chore into a major source of business intelligence.   They cover everything from decreasing costs to improving accuracy and enhancing customer support.   We also hear from Fouad Farhat of Soma Capital—an investor in both Truewind and Puzzle—about the current state of accounting, and where it’s headed.   📌Check out the full article here: https://lnkd.in/eUzvZjTm

22

Some thoughts on being available "off hours": I've seen a few threads on LI/otherwise about being available for work during off hours, vs having work/life separation, and changing trends there (ie: younger people are less willing to let work intrude in their personal time). As someone who has been in both situations in his career, I have some thoughts. A company will always want employees to work 24/7 (at least exempt employees). They will applaud employees who "give extra", "do what's necessary", "are good team members", or whatever other euphemisms are deemed effective to encourage this. But, employees are also generally disposable to employers (especially larger companies). This is often a rude awakening for younger people, who have been conditioned with propaganda like "your work family", etc. Generally, you should work for compensation, not some sense of obligation to a company that would discard you without a second thought for a small bump to their bottom line. Ideally, those types of expectations should be up-front and by consensus with one's employer. Generally, circumstances permitting, I will check messages off-hours, and respond/work if there's a critical situation which needs my attention (and not otherwise). However, that's inherently dependent on the employer also; if the employer makes that "hard" (eg: restricting the ability to check messages to only corporate devices), then I interpret that as a signal that working off hours is not something they value (at least enough to make it "easy"), which is fine with me. In some ways, that's a better situation, if/when the employer is implicitly making it hard to work off-hours anyway, it easier to ignore any implied pressure to do so. Anyway, that's my take, fwiw.

3

𝗔𝗿𝗲 𝘆𝗼𝘂 𝗹𝗼𝘀𝗶𝗻𝗴 𝘀𝗹𝗲𝗲𝗽 💤 ? because you can't pinpoint your CI bottlenecks or spot problematic trends? At Pinterest, we faced the same challenges—but we’ve found a solution. 🌟 In our latest Pinterest Engineering blog post 🔗 https://lnkd.in/eJJ_GpGW We share how we’re leveraging honeycomb.io to identify CI root causes through cross-data correlations. Discover our innovative approach for categorizing CI failures in real-time using job log analysis. Learn how to segment CI jobs by agent wait time, script execution, specific Bazel targets, or any other level of detail tailored to your needs. Our enhanced CI observability metrics have let us improve CI stability and boost developer productivity—and they could do the same for you! Check out the blog for the full breakdown and see what’s possible when data-driven decisions guide your CI workflows. 🚀 Let’s make CI faster and more reliable together! #DevOps #DevX #CICD #honeycomb #observability #engineering #softwareengineer

16

2 Comments

A data-driven approach to driving wound healing outcomes Marcella Gomez, University of California, Santa Cruz https://lnkd.in/gAkWZYJT Meet: https://lnkd.in/g4AAWkcK YouTube Stream: https://lnkd.in/gwtqaTmf Talk details: https://lnkd.in/gWn3B3cU Subscribe for talk announcements: https://lnkd.in/g5ciuNuX Abstract: Precision medicine requires an ability to predict the response of an individual to a prescribed treatment regimen a priori. Thus, advancement in the field is challenged by a lack of predictive models and, arguably, a lack of time-series information for a highly dynamic system. Here, we discuss work in wound healing, for accelerating wound closure. We argue that timing of treatments is as critical to consider as the choice of drug or therapy. Due to system size and complexity, data-driven methods need to be explored to develop multi-dimensional quantifiable indicators tracking systemic changes. In this work I discuss how bioelectronic devices enhanced with deep learning can help facilitate real-time sensing and actuation for automated decisions in treatment for wound healing and preliminary work in transcriptomic based classification of wound states. Bio: Marcella M. Gomez is an associate professor at UC Santa Cruz in the department of Applied Mathematics and Associate Dean for Diversity, Equity, and Inclusion for Baskin Engineering. She received her PhD from Caltech in 2015 and a B.S. from UC Berkeley in 2009; both degrees in Mechanical Engineering. Her research interests are in the broad field of bio-control leveraging methods in machine learning and control theory. Applications range from controlling single-cell response to driving complex systems such as wound healing. #modeling #simulation #medicine #healing #trauma #predictivemodeling #datascience #data #woundhealing #biology #bioengineering #systemanalysis #sensing #omics

3

Wow. So, in terms of LLM observability & prompt management/optimization there's so much out there: Freeplay (Colorado yay!) LangChain's Langsmith Weights & Biases's Weave Langtrace Langfuse (YC W23) [Edit] Added from comments / messages: LangWatch orq.ai https://lnkd.in/gFATt6mn Humanloop

21

3 Comments

We're in the pets phase of LLMs right now, but the cattle phase is coming and it'll change everything. Yes, again. What do you mean "pets" and "cattle"? Servers used to be our pets. We bought, named, cared for, and persisted them individually. If we wanted a new one, it was a big commitment. If one went down, it was a big loss. But then servers became cattle. Today we manage them as a group of commodity assets. We want one, we clone it. If one goes down, we replace it with another. This change delivered an era of distributed, high availability applications with limited downtime. Cheap storage, containers, cloud, clusters all met riding on top of Moore's law. And machine learning changed too. Computer vision was first. We went from small, neural networks running on CPUs via Matlab to a larger networks running in on GPUs via C++. AlexNet kicked off the deep learning revolution, eventually Transformers in NLP, and then ChatGPT. So, today we call our LLMs by name but this is just the start. There will be more LLMs variations than there are people on Earth.

25

1 Comment

Check out this conversation I had with Amir Bormand on The Tech Trek Podcast on effective management and how individual contributors can leverage their manager to move their careers forward

85

12 Comments

Migration Complete & Holiday Success 🎉 What seemed like a monumental risk—migrating an entire platform in just over two months, right before launching our most crucial holiday promotion—could have easily been a recipe for disaster. But at Caraway Home, we don’t shy away from big challenges. Instead, we rally around them 🚀 I also want to state how incredibly grateful I am to Jordan Nathan and Connor Dault for trusting in me to lead the charge. Their confidence in my ability to steer this project was the key to our success. We dove into re-platforming our website with one goal: to modernize, scale, and prepare for the future. Every part of this migration had to be seamless—performance, speed, and scalability had to be perfect for the holiday rush, and we had zero room for error. It was a team effort that required every member to step up and give their absolute best. Using Shopify's Hydrogen-React, Meta Facebook's Stylex-JS, Next.js, TypeScript, Vercel, and Contentful, we pulled it off. The result? An incredibly fast, flexible, and high-performing site that not only supported our holiday promotion flawlessly but set us up for continued growth in the future. We were also able to reduce build times by 80%, skyrocketed organic search sessions, and delivered a smooth, seamless experience for our customers. 🌟 Our page speed improved 2x, ensuring fast load times even during peak traffic. We also maintained an average Core Web Vitals score of 90 🤯 throughout the entire holiday promotion. Engagement also saw a significant boost, with more customers interacting with our site and exploring our products than ever before. Additionally, we introduced a more cohesive style guide and design system, making the developer & user experience not only faster but also more visually unified and on-brand across all touchpoints. ✅ Multiple $ Record Breaking days during the holiday season 💰 ✅ A site built for speed and scalability ✅ Improved developer experience (editors are loving it too!) ✅ A superior customer experience This migration wasn’t just a technical achievement—it was a bold move that shows what’s possible when you have the right team and the right tools. I’m incredibly proud of what we’ve built over the last five years at Caraway, and I’m beyond excited about what lies ahead. Huge thanks to everyone on our team who made it possible Alexa Bravo-Fuentes, Kelsey Yin, Ethan Shaffer, Katie Krobock, Ashley Golladay, Emily Mercado, Ariel Rodriguez, Dawid Urbański, Will Lee, Neeraj K., Dellean, Juan, and the entire team for making this possible. This was truly a group effort. If you haven’t checked out the new site yet, now’s the time! https://lnkd.in/e8MQ7RyC Shop our best sellers and holiday picks, and save up to 20%. 🎁 #eCommerce #Leadership #Caraway #Nextjs #ShopifyHydrogen #HolidaySuccess #stylexjs #meta #shopify #headless #react

87

30 Comments

Anthropic released another paper on their ongoing work towards interpretability of large foundational models like their Claude 3 family, built on previous work by many. Their work continues to be enlightening and inspiring. The findings in this work are generally similar to previous findings on smaller models but scaled up to Claude 3 Sonnet: 💡 they find large numbers of "features" corresponding to combinations of neuron activations and these features map to concepts that we recognize/understand: for example, features for transit infrastructure, particular people, concepts (like bias and "objects lacking sentience"). This might sound familiar from classical machine learning, but the interesting thing is that there is no single thing/neuron encoding each of these concepts, it's a particular combination of neuron activations that embeds these concepts in "superposition" in the network. 💡they observe a sort of "distance" of these feature activations that arrange themselves "near" semantically similar features. When they use their sparse autoencoder approach to learn more and more features they observe that the level of abstraction in the detected features becomes more granular, in ways that are interpretable and meaningful. 💡when they artificially fire these activation patterns during inference, it affects the output in ways that relate to the "feature". For example, asking Claude "what's the most interesting science" normally it responds Physics, but if you clamp the feature activations for the feature "Brain Science" to 10x their max values then Claude responds: Neuroscience. Ways to interpret _and_ influence the model. Great readable work [1] along with an interactive explorer of some of the "features" themselves [2] 🎂 There's an easter egg too: if you scroll in the feature explorer tool and try and select one of the "Randomly selected features from 1M" you get a message "if you want to see the rest of the features, we're hiring!" with a link to their recruiting page :) [1] https://lnkd.in/gZbparwp [2] https://lnkd.in/gC9eus53

19

After numerous recent conversations with engineering leaders, it's clearer than ever to me that when it comes to software development, we're caught in a reactive app performance management cycle. Teams prioritize rapid feature shipping, inadvertently accumulating performance debt with each release. The pattern is frustratingly predictable: 1. Ship features quickly 2. Performance gradually degrades 3. Users complain or infrastructure costs spike 4. Briefly optimize performance 5. Repeat the cycle This approach is fundamentally broken. Performance ideally should be a first-class citizen in our development process, not an afterthought triggered by crisis. With Codeflash, I'm exploring a more proactive approach to performance engineering. The goal is to build performance consciousness into our development workflows, preventing technical debt before it becomes a systemic problem. I'm eager to hear from other engineers: How are you addressing performance challenges in your organizations? What strategies have you found effective in maintaining software efficiency without sacrificing development speed?

30

1 Comment

I keep reading this every couple days and falling in love with it more. We keep having more and more systems because people need to write articles and books. And, of course, you have to have your own acronym. But when we step back and look at the core of things, most of it boils down to: 1. Go to therapy and deal with your own shit 2. Talk to other people. Be humble. 3. Actively listen to them and think deeply about what you heard from them 4. Don't be a dick. Have empathy. 5. Try to figure out situations where everyone wins That's it.

7

1 Comment

When Klaviyo recruited me three years ago to help architect a system that would deliver hundreds of millions of email messages per hour on behalf of their customers, I thought, "Wow. That's a lot of mail. How could I *not* build that?" Now that we're doing it for real, I still think, "Wow. That's a lot of mail. We *built* that." $3 billion+ in Klaviyo-attributed value. 420 million messages sent in a single hour on #BlackFriday. And over 15,000 customers having their best day ever (🙌) on Klaviyo.  Read our full BFCM By The Numbers report to learn how Klaviyo customers achieved recording-breaking results this #BFCM.

35

9 Comments

The CIDR paper I co-wrote and mentioned a couple months ago is now available for the public to read. Take a look and let me know if you have any feedback! https://lnkd.in/gVkcasRz

121

2 Comments

While some folks hire entire Figma design teams for their personal websites, I'm here at 6:30 AM explaining to my partner why I need a third coffee to "write stuff." 2022: "I'll share monthly insights about building companies!" 2024: checking it -> Writing consistency matches crypto winter performance 📉 But look - AI got better at English faster than I got better at writing. That's what we call product-market fit. New strategy for 2025 - Actually write 12 articles - Convince partner these dawn coffee raids are essential business ops Track my progress (or lack thereof) here: https://lnkd.in/e6bkkxvH P.S. If my partner reads this - yes, this is exactly what I'm doing at 6:00 AM either wiring or reading the book. Note: Will open source my coffee consumption metrics in Q3 2025. Full transparency, like web3 promised.

30

1 Comment