Toby Jaroslaw

I've been thinking a lot about the rise of #generativeAI, and how it reminds me of some of the things I've encountered in my background as a scholar of ancient texts. Here's a piece about some concerns about over-simplifying the things that AI offers in the world of tech. #blog #academic #translation #software #coding https://lnkd.in/ePT3Qcrs

11

I've subscribed to dozens of newsletters to stay up-to-date about climate change. HEATED (https://heated.world/) by Emily Atkin is by far the most interesting to me. It's the only one I pay for each month. Emily inspires and informs where others only provide data and facts. She keeps the content up-beat while not downplaying the rage-inducing actions of certain organizations and leaders. I didn't realize how much effort these leaders were putting in to create false narratives that distract the public from their destruction of the environment. I knew it was happening, but I didn't realize how often and to what lengths they were going to keep the public misinformed. These leaders want to keep selling product, expanding operations and making money at the expense of... their livelihoods. I can't blame them for wanting to make more money, but do these people not have children they care about? Do they not see the connection between their actions and the data regarding climate change? I'd love to know because it just seems so obvious. I recommend subscribing to her newsletter and supporting her journalism. She's doing some of the most important work of our time. I guarantee you'll learn something new. If this post only convinces one person to subscribe, I'll consider it a success. It won't happen overnight but I believe education is the way out of this mess.

7

Share your story about Ray at the Ray Summit! I'm happy to chat about your talk abstract if you're building your platform with KubeRay.

14

2 Comments

🚀 Exciting Announcement: Launching MergeMate! 🚀 After many weekends of dedicated effort, I'm thrilled to unveil a project that's close to my heart: MergeMate. It's a robust GitHub action that leverages advanced AI to automate pull request reviews and command handling, significantly enhancing workflow efficiency on GitHub. 🔗 GitHub: https://lnkd.in/grnkju3j MergeMate is not just another tool; it's an intelligent assistant designed to streamline the way we manage code reviews and interact within GitHub projects. By integrating AI capabilities, it provides context-aware feedback, handles interactive commands, and ensures that best practices are always at the forefront of our development processes. 🎥 To see MergeMate in action, check out our demo on YouTube: https://lnkd.in/gZjYcPTn 🌟 If you find MergeMate interesting or useful, please consider giving it a star on GitHub! Your support is invaluable and helps us reach more developers. 🔍 Open to Contributions & Issues: The future of MergeMate includes expanding to platforms like Bitbucket and GitLab. I'm open to contributions and would love to address any issues you face. Let’s collaborate to make development easier for everyone! I'd love to hear your thoughts and feedback on MergeMate. If you're as passionate about improving development workflows as I am, let's connect and discuss how MergeMate can be part of your toolchain! #GitHub #AI #DevTools #OpenSource #MergeMate #SoftwareDevelopment

28

5 Comments

🚀 **Cracking the Anagram Code: Insights from Headstarter** 🚀 Hi LinkedIn network! 🌟 I recently tackled a compelling coding challenge at Headstarter AI: verifying if two strings are anagrams. Here’s a breakdown of my approach and insights: 🔍 Challenge Overview: The goal was to determine if two strings are anagrams—essentially, checking if they contain the exact same characters in any order. 🔄 💡 My Strategy: - Length Check: First, I verified that both strings are of the same length. If they’re not, they can’t be anagrams. 📏    - Character Counting:  - Utilized two dictionaries to track the frequency of each character in both strings.  - Updated counts for characters in the first string and adjusted counts for the second string. 📊 - Validation: After counting characters, compared the two dictionaries. If they match, the strings are anagrams! ✅ 🌟 Why This Approach? - Efficiency: Ensures linear time complexity, making it effective for large datasets. ⏱️ - Minimal Space Use: Uses dictionaries efficiently, keeping memory usage low. 💾 - Simplicity: Provides a straightforward and clear solution. 🧩 🔧 Key Learnings: - Mastering data structures and algorithms is essential for solving complex problems. - Often, the simplest solutions are the most powerful. 🚀 🙌 Gratitude: A huge thanks to the Headstarter team for these thought-provoking challenges! 👏 🌐 Join the Discussion: How do you tackle similar coding challenges? Share your strategies and experiences in the comments! ⬇️ #CodingChallenge #SoftwareDevelopment #AlgorithmDesign #AnagramSolution #HeadstarterAI #TechCommunity #ProblemSolving #Innovation #LearningPath #TechInsights #LinkedInTech

1

1 Comment

If you're developing with FastAPI, you might find this handy: I've written up how I setup database connections with the framework. https://lnkd.in/eEBnuW7z

12

2 Comments

I wrote about about the state of compute in the world of AI and what that means for individuals. I also included my usual dump of ML resources and updates in a streamlined fashion. I'm changing up the format of my writing this week. Let me know what you think! https://lnkd.in/dS_EXsK5

5

2 Comments

Is AI a new commodity infrastructure, or is it a source of new platforms and products? The recent launches of Apple and Figma “AI” have made many people think about the actual role of AI (or LLMs) in products. If you zoom out from their newest demos and look at their strategy with AI, you see they go against the hype of general-purpose chatbots that can solve and automate various complex tasks in a single interface. Instead, they “unbundle AI” into many small features and places, mostly invisible or hard to see directly. This highlights two distinct approaches to building with LLMs right now. On the one hand, some are bundling various use cases into new interfaces with LLMs at the core, creating entirely new working experiences. On the other hand, others are embedding LLMs into specific, familiar interfaces, enhancing existing workflows in a more seamless way. There are strong advocates for both sides of the spectrum. But for me, I think, “Why not both?” Because the two actually have different purposes and experiences, they work best for different cases. General-purpose interfaces are good at iterative processes but lack structure, so they may work best for tasks that are simple and independent. While specific UIs have structure but lack flexibility, making them more suitable for tasks that require fixed processes and dependencies. One main reason behind this divide, I think, is not from the LLM's ability itself but more from the “starting point” of companies building on top of it. Existing companies and incumbents want to integrate AI into their existing products and turn them into features, while startups try to reinvent the wheel and create something new from the ground up. This is not something new in the tech industry, we saw it all the time before AI. The only difference now is how AI has the potential to do most tasks we know, really well, and all in one place. Any thoughts?

5

Let’s talk about “dopamine culture” in #digitalhealth 🧠💻 I’ve recently read an article on the concept of “dopamine culture”. It got me thinking of my own addiction to what they call “instant gratification”. As a student eager to make my mark, the allure of rapid, tech-driven solutions was hard to resist. Yet, the dopamine rush of quick fixes often glossed over deeper health needs (mine included). The ease of digital tools seemed to widen the gap between genuine healthcare interactions and the superficial engagements that apps offered. I saw peers swipe through health advice with the same casualness they'd use on dating apps, chasing the dopamine high of instant solutions. The consequences? A generation potentially detached from the nuanced understanding of health, where the touch of a button could override the touch of a hand. So, is tech all that bad? Of course not! Health tech is a blessing for our modern day health disperaties. It can offer people equitable health from all societal layers. However, everything in life is good in moderation. We need to integrate tech to extract it’s undeniable benefits but grounding our human-centered care. How can we marry convenience with compassion? 👉 Developing tech that incorporates inclusive health outcomes rather than narrow commercial interests. 👉 Developing tech that prioritizes long-term outcomes over short-term engagement metrics. 👉 Digital literacy to differentiate between quantity and quality. As Spencer Dorn said “If we try to speed up too much, we devalue the critical test of time. And we tarnish (or discard) the therapeutic relationships and time-honored clinical rituals that facilitate healing.” Do you find yourself falling in the dopamine culture pitfall? Share your experiences in the comments 👇💬 🔗 https://bit.ly/4aBVwaR —————————————————— 𝘏𝘪! 𝘐 𝘢𝘮 𝘎𝘩𝘢𝘥𝘦𝘦𝘳 𝘈𝘭 𝘙𝘶𝘸𝘢𝘪𝘴𝘩𝘦𝘥𝘪 👋🏻 𝘈 𝘉𝘪𝘰𝘮𝘦𝘥𝘪𝘤𝘢𝘭 𝘌𝘯𝘨𝘪𝘯𝘦𝘦𝘳𝘪𝘯𝘨 𝘴𝘵𝘶𝘥𝘦𝘯𝘵 𝘸𝘪𝘵𝘩 𝘢 𝘱𝘢𝘴𝘴𝘪𝘰𝘯 𝘧𝘰𝘳 𝘈𝘐 𝘪𝘯 𝘩𝘦𝘢𝘭𝘵𝘩𝘤𝘢𝘳𝘦 🤓 𝘐 𝘸𝘳𝘪𝘵𝘦 𝘱𝘰𝘴𝘵𝘴 𝘦𝘷𝘦𝘳𝘺 𝘔𝘰𝘯𝘥𝘢𝘺 𝘢𝘯𝘥 𝘛𝘩𝘶𝘳𝘴𝘥𝘢𝘺 𝘢𝘥𝘷𝘰𝘤𝘢𝘵𝘪𝘯𝘨 𝘧𝘰𝘳 𝘣𝘦𝘵𝘵𝘦𝘳 𝘩𝘦𝘢𝘭𝘵𝘩 𝘦𝘲𝘶𝘪𝘵𝘺 𝘢𝘯𝘥 𝘳𝘦𝘴𝘱𝘰𝘯𝘴𝘪𝘣𝘭𝘦 𝘈𝘐 𝘪𝘯 𝘩𝘦𝘢𝘭𝘵𝘩𝘤𝘢𝘳𝘦 🔎🩺 𝘓𝘦𝘵’𝘴 𝘤𝘰𝘯𝘯𝘦𝘤𝘵 𝘢𝘯𝘥 𝘭𝘦𝘢𝘳𝘯 𝘧𝘳𝘰𝘮 𝘰𝘯𝘦 𝘢𝘯𝘰𝘵𝘩𝘦𝘳! 🔗 —————————————————— #medtech #healthtech #healthcareonlinkedin #healthcare #aiinhealthcare #aiethics #patientcentricity #healthcareinnovation

21

5 Comments

As the AI hype cycle continues, you're also starting to see the "data is the new oil" make its way back into the zeitgeist. It's undeniable that data is valuable, especially proprietary datasets, but how valuable? Up until recently there weren't data marketplaces so it was difficult to gauge how the market would value a dataset. But now with the Snowflake, and other, data marketplaces we can get a sense of what people are willing to pay for data. If you check out the marketplace another interesting aspect is how many datasets are packaged up public data. For example most of the weather data, https://lnkd.in/euhS4GbB would be available for free from NOAA. We've been experimenting with this as well. We've aggregated all the public data from Dutchie eCommerce sites and made it available on Snowflake, https://lnkd.in/eJRTZd5S BDSA sells a full product based on this dataset, https://lnkd.in/ejMqCWuc Broadly, it feels like there's some interesting opportunities here to aggregate, normalize, and sell public data. Have you used any public datasets that are re-packed into a paid product?

2

Week 1 Update - Headstarter AI Fellowship 🚀 This post is dedicated to sharing some impactful highlights from my first week at the Headstarter AI fellowship Personal Portfolio Website 🌐: After months of working tirelessly, to create a personal portfolio I struggled with either using an HTML5 template or just making the backbone of my website written in pure HTML, CSS and Javascript. I finally came to the conclusion that the best solution was to implement an HTML5 template since it would come with basic Javascript functions needed for my website to work. I have now deployed it on Netlify and I am now working on fixing the bugs I have on my Github pages. I will be sharing the link 🔗 with everyone after I finish touching up some edges! Networking with Engineers 👨‍🔬 🤝 : This past week, I had the rare opportunity of connecting with more than 5 professionals that have had great impacts within their own diverse fields and it was quite satisfying when they answered my questions with great insights and enthusiasm! Mock Interview 💻 : I completed my first DSA mock interview with Headstarter's AI version of Aidan Ouckama a current SWE intern at DoorDash, and it was such a phenomenal experience - quite challenging but also satisfying at the same time! Key Takeaways: 1. I have learnt that the human mind is one of the most powerful tool within our possession and how or what we think really matters. Therefore, it is important that we open our minds to constantly learning and also failing in that process. With this I am very excited to learn how React and TypeScript works with mobile apps development process! Stay tuned for more updates in the coming weeks! 2. Team Collaboration: Today I met with my teammates to brainstorm on what we would be doing for the Headstarter AI projects and also discussed our strategies for our very first Hackathon! From this session, I have learnt how vital it is to collaborate with other people in Tech! I am looking forward to what my teams and I will produce! #HeadstarterAI #Fellowship #Networking #CareerNetworking #ComputerScience #SoftwareEngineering #PersonalWebsite

16

1 Comment

It’s about time I share! I founded an AI startup 🤓 With Pandria.com we’re tackling a familiar problem: You’ve likely felt the collective annoyed sigh when it’s time for the semi-annual performance review cycle. You log into some system, you push buttons, people get reminders to fill in forms about you. Everyone waits until the last minute, then rage-types haphazard insights from the past two weeks. The resulting ‘growth conversation’ is generic at best, usually awkward, and biased at worst. That culture of ongoing, actionable, constructive feedback that is supposed to feed professional growth lives only in the wishful minds at the People department. I think we’ve cracked this 😀 We’re onboarding partners keen to test our solution, so I'd love to tell you more! (In fact, I can’t shut up about this - ask anyone I know.) So go ahead, give Pandria a follow, and do reach out if this problem strikes a chord. Excited to introduce the all-star band soon 🔥

157

35 Comments

The biggest problem in the LLM market right now is people talking about or influencing people to think about cost-quality trade off. If I don’t even know whether AI will be worth trillion dollars or million dollars what’s the point in minimizing cost per token? I don’t think we’ve explored conceptually what LLMs are capable of. Actually I don’t think we’ve conceptually understood AI yet. I suspect we will have another stock boom with AI companies like the 2000 stock bubble burst for the Internet. Then the hardy ones that survive that will continue to make really amazing LLM products, much much better than the ones we have now. You’ll know it. Once the market is really ripe for AI you will suddenly get a sense that people around you are talking about how to leverage AI rather than how to leverage the hype of AI. It’s important to notice this subtle shift in narrative and hopefully pounce on the right investments. Yes it would be weird to talk to your AI girlfriend and suddenly hear it talk about having a cool and refreshing beverage like “Coca cola!” But hey. We already don’t know how many YouTubers and other Influencers are nudging us towards certain brands/topics.

8

1 Comment

LLMs: Redefining What We Mean by "Application" I've been wondering how Large Language Models (LLMs) might fundamentally alter our concept of software applications. 1. From Rigid Logic to Fluid Reasoning: Traditional apps operate on predefined rules. LLMs, however, can "reason" across vast knowledge domains. Could this shift from if-then statements to contextual understanding redefine app functionality? 2. The Universal Interface: Imagine an app that's essentially just an interface to an AI. You describe what you want, and it figures out how to do it. No menus, no predefined functions - just natural language interaction. Is this the future of user experience? 3. Adaptive Functionality: Current apps have fixed features. But what if an app could dynamically generate new functionalities based on user requests? LLMs could potentially turn every app into a swiss army knife of capabilities. 4. Personalization at the Core: LLMs can tailor responses to individual users. Could this lead to apps that fundamentally reshape themselves for each user, rather than just tweaking surface-level preferences? 5. The End of Updates?: If apps become interfaces to continuously learning AIs, do we move beyond the concept of versions and updates? The app might evolve organically with use. These possibilities challenge our very definition of what an application is. Are we moving towards a world where the distinction between different apps blurs, replaced by personalized AI interfaces that can do... well, anything? What do you think? Are LLMs set to revolutionize not just how we develop software, but how we conceptualize it entirely? How might this shift impact software development, UX design, and even digital literacy? #FutureOfSoftware #LLMs #AIInterface #TechEvolution

1

I challenged AI to solve a personal medical problem for me, and it did well, so I thought I'd share: I've done several blood lab tests a few weeks apart to monitor specific indicators. The results were provided in PDF files with varying file structures, metric names, and acceptable ranges across different labs. I needed to compare them and see how the metrics perform over time. Classic BI problem, right?! So here's what I did: Uploaded lab results PDFs to Claude 3 Opus with the following prompt: """Here are blood test results. Export each metric and its value into CSV with the following fields: 1) Test name. Use canonical test names, e.g. if a test is called "WBC" in one file but "White blood cell count" in another, use the same name for both. 2) One-line description for the test (canonical) 3) Test value 4) Test unit (canonical) 5) Date the test was collected 6) Normal range (canonical) 7) TRUE/FALSE flag for whether or not the result is in range or out of range""" I was surprised with the quality of the output: 1 - Claude correctly transcribed the results 2 - it did well with standardizing test names and ranges by taking the min of the lower range and max of the upper range (reasonable!) 3 - it produced a clean CSV I then uploaded the CSV into a spreadsheet and plotted it over time with a pivot table. All in all, it took ~10 minutes and would take hours to do manually. The same workflow is possible with GPT-4 but it threw errors when I tried to generate CSVs, so props to Anthropic for having a lead here!

64

Doctors hate charting. I built a tool to automatically generate structured clinical notes from an audio memo - or a recording of the entire appointment. A workflow for many independent clinicians making house calls is to do an audio memo, send this to someone for transcription, get the drafted note back a day or so later later, and then make final edits, sign, and upload to a secure database. This cuts the process down considerably: go from audio memo / appointment recording to structured SOAP note in about a minute. Review and sign. It works surprisingly well! I was even able to tuck in an additional feature: the app automatically generates a "second opinion" for the clinician to consider. Potentially decreases the possibility of clinician error or oversight. Built with Replicate and Supabase. Deployed with Vercel. Currently doing evals with models from Anthropic, OpenAI, and Gemini AI #medtech #healthtech #healthcare #ai

16

5 Comments

Yesterday I completed some work with a JPMorganChase project through Forage. The task had me work with JPMorgan Chase's Perspective software. Perspective is an open-source interactive analytics and data visualization component, which is especially well-suited for large and/or streaming datasets (https://lnkd.in/ebzXQXwU), (https://lnkd.in/eduSjb9k). This was a great project because I had the opportunity to work with Python's Unittest framework, which I seldom use. I also was able to work with creating a git patch file, which I often have never had to use in my current role. The utility of being able to upload patch notes specifically to a service-ticket is awesome, and I can definitely see myself continue to use this tooling for future personal and professional projects. Today I am going to continue working on the project by working on the frontend portion of Perspective, and testing my skills with Typescript (I love Typescript). Should be fun! If you have not checked out this project on TheForage.com, or Perspective by JPMorgan Chase, I recommend it! #SoftwareEngineer #JPMorganChase #Perspective #Developer #Python #Git #Github #LookingForWork #OpenToHire #backend #frontend #Typescript

6

My thoughts after reading "Tidy First?" by Kent Beck (O’Reilly, 2023). In a few words, the book is unusual and surprising. It's quite short – less than 100 pages, almost all chapters are just 1-2 pages long. The language is simple, it's easy and enjoyable to read. It starts with simple things and ends with something that can completely blow your mind! To me, the book is a condensed piece of wisdom shared with us by someone who has been in the industry for decades. As you can see from the subtitle, "A Personal Exercise in Empirical Software Design", the book offers a novel point of view on software design, and it focuses on you as a developer, on what you can do to improve design during your day-to-day work. It starts with a set of “tidyings”, which essentially are micro-refactorings. These tidyings make the code more readable, easier to comprehend and change. While there might not be anything entirely new, it's a great set of simple practices that make a developer’s life easier. It's extremely valuable for junior roles and a great refresher for seniors. Next, the author talks about when to do (or not to do) those tidyings, how to combine them together and with other kinds of changes. This section offers plenty of food for thought and ideas for improving processes within your team. Valuable for any developer and maybe even managers. The last part is philosophical, and I found it the most intriguing. It provides completely unexpected perspectives on software design! Well-known things like coupling and cohesion are combined with unique ideas on software development economics and optionality. Must read for seasoned professionals. Also, the end of the book answers the question asked in the title - Tidy First? The author says that this book is the first one in the series, aimed at individual developers. The next book will target the team level, followed by one on business level matters. Thank you, Kent, for a great read during my vacation, and looking forward to the next book! https://lnkd.in/ekyRx9DC

5