Krishnan Sriram

Important to use the right word for the right situation

4

Very recently, I had the opportunity to cook for over 70 people, a challenging experience showcasing the complexity of scaling up simple tasks. What appeared effortless on a small scale became intricate with larger volumes, mirroring challenges in platform architecture and automation. At low volumes, imperfections may slip by unnoticed, but scaling up demands perfection. The key is to emulate a production line - meticulous organization, understanding, and preparation are paramount for seamless operations at scale. Lessons learned? Firstly, experimentation at scale is a no-go; perfect first, then scale. Secondly, dont under estimating scaling complexity. Lastly, respect those who excel in the field; it's truly an impressive feat. #PlatformArchitecture #Automation #ScalingUp r

19

7 Comments

Super happy for Greptile and for Daksh Gupta for raising $4M to build AI-Powered code base expert. I had the opportunity to speak with Daksh, he walked me through how Greptile works. Check out the full video here #genai #greptile #codeassist

17

Just finished Microsoft Power Apps: Using the Dataverse (Formerly the Common Data Service)! Check it out: https://lnkd.in/geRkeWEd

24

# AI Testing Challenges 1. *Garbage In, Garbage Out!* Bad data = bad results. Feed your AI testing tool junk, and it'll spit out junk. 2. *AI with an Attitude!* Biased AI = missed test cases. If your AI testing tool has an attitude, it might ignore important test cases. 3. *Magic 8-Ball Testing!* No explanations = confusion. If your AI testing tool doesn't explain its results, you'll be left wondering what's going on. 4. *Sneaky Hackers!* Insecure AI = messed-up results. If your AI testing tool isn't secure, sneaky hackers might mess with your results. 5. *Puzzle Piece Problem!* Integration issues = frustration. If your AI testing tool doesn't fit with your existing testing setup, it's like trying to force a puzzle piece into the wrong spot. 6. *High-Maintenance Car!* Too much maintenance = pain. If your AI testing tool requires too much upkeep, it's like owning a car that's always breaking down.

1

This year's Biztalk replacement project is finally complete, we've replaced our legacy interfaces with shiny new ones. This made me think of the resiliency of the architecture and design of the original solution, and the 𝘱𝘦𝘰𝘱𝘭𝘦 who did it. The legacy services were designed in 2011, upgraded in 2013, and have only been updated 5-6 since then in the few projects that funded required changes in them. But the have performed flawlessly handling thousands of transactions a day with a 99.999% uptime and reliability score. These services have integrated several enterprise systems, platforms and workflow that have the same design patterns and legacy. They are 6-20 years old and have had minimal maintenance and support.  "You must have great infrastructure to support such a collection of legacy systems". No, we have great systems running on the fewest resources. There's no load-balancing for these 70 applications hosted on 3 unique servers. No problem. 𝟵𝟵.𝟵𝟵𝟵% 𝘂𝗽𝘁𝗶𝗺𝗲. No system events. No memory faults. No Code exceptions. No Data faults. No integration failures. All rock-solid applications still running the same base code that was designed long ago using the popular pattern and tools of the time. 𝗜'𝘃𝗲 𝘄𝗼𝗿𝗸𝗲𝗱 𝘄𝗶𝘁𝗵 𝘀𝗼𝗺𝗲 𝗲𝘅𝗰𝗲𝗽𝘁𝗶𝗼𝗻𝗮𝗹 𝗽𝗲𝗼𝗽𝗹𝗲 𝘄𝗵𝗼 𝗵𝗮𝘃𝗲 𝗱𝗲𝘀𝗶𝗴𝗻𝗲𝗱 𝘀𝗼𝗺𝗲 𝗳𝗮𝗻𝘁𝗮𝘀𝘁𝗶𝗰 𝘁𝗵𝗶𝗻𝗴𝘀. 𝗜 𝗵𝗼𝗽𝗲 𝘁𝗵𝗮𝘁 𝗶𝗳 𝘆𝗼𝘂 𝗿𝗲𝗮𝗱 𝘁𝗵𝗶𝘀, 𝘆𝗼𝘂 𝗮𝗿𝗲 𝗼𝗻𝗲 𝗼𝗳 𝘁𝗵𝗼𝘀𝗲 𝗽𝗲𝗼𝗽𝗹𝗲 𝗮𝗻𝗱 𝗸𝗻𝗼𝘄 𝘁𝗵𝗮𝘁 𝗜 𝗮𝗽𝗽𝗿𝗲𝗰𝗶𝗮𝘁𝗲 𝘁𝗵𝗲 𝗮𝗺𝗮𝘇𝗶𝗻𝗴 𝘄𝗼𝗿𝗸 𝘆𝗼𝘂 𝗱𝗶𝗱.

36

23 Comments

I’m happy to share that as part of my #GenAI learning I’ve earned my 'Inspect Rich Documents with Gemini Multimodality and Multimodal RAG' skill badge. #GoogleCloudLearning #GoogleCloudSkillBadge #Gemini #Multimodality #RAG #VertexAI #GenAI

8

A Remarkable Improvement in Developing Software with Claude 3.5 Sonnet. Leonardo da Vinci wrote that a poet would take a very long time to "describe with words what a painter is able to [show] in an instant." The same can be said for a developer typing text into a Gen AI chat compared to showing a screenshot to that AI. I just discovered Claude 3.5 Sonnet's screenshot analysis and how useful it is for the visual side of software. This fundamentally changes the collaborative discussion between the developer and the AI. The ability to instantly analyze, understand, and extract information from UI screenshots is a much more intuitive way for resolving issues where some screen element that used to work suddenly disappears. Combined with Claude's project feature, I can load its context window with thousands of lines of code in my project, discuss with Claude how to simplify and improve a specific file, and then use both text and screenshots to discuss the results of its suggestions. This is very powerful. It has accelerated my ability to quickly code complex software projects by: - Quickly analyzing error messages and UI states from screenshots instead of requiring me to type in a description of what I was seeing - Automatically identifying visual inconsistencies - Providing a common context to discuss the visual experience with Claude, because the AI and I both visually see the same thing, instead of relying on text descriptions Shared visual context also extends to accessibility and ADA compliance. Claude has the dual ability to view both the screen and how it is semantically structured in the code. This makes it an invaluable tool for ensuring accessibility standards are met and for improving the experience for all users. #AI #SoftwareDevelopment #DevTools #CodingProductivity #TechInnovation #Claude #AnthropicAI

2

1 Comment

Webinar: Successfully apply AI for application modernization Generative AI has the power to transform the process of Application Modernization through code reverse engineering, code generation, code conversion from one language to another, defining modernization workflow, and other automated processes. This webinar is designed to equip technology leaders with the knowledge and understanding necessary to evaluate your technology estate, assess the opportunity and chart a path forward. https://lnkd.in/e8NybGTW

8

My teammate Swara Gandhi just released the Resource Control Policy (RCP) example repo live to github! This is a great place to get started with some example RCPs, go check it out! We accept pull requests, by the way, if you've got some examples you'd like to share :) https://lnkd.in/gWmuy9sd

93

Diving into AutoGen has been a great experience! This course is highly recommended for anyone looking to learn more about architecting and building autonomous agents. #artificialintelligence #autogen #autonomousagents

9

Govern your software architecture

1

The world of FLEXCUBE lost one of its great unsung heroes yesterday. Radhakrishna Murthy is no more. Radha, as he was fondly called, will forever be remembered by the FLEXCUBE community for his visionary creation—the RAD tool, formally known as FLEXCUBE Open Development Tools (ODT). Radha was synonymous with the tooling; so much so, that in the early days, some even believed it was called "RAD" because RADha (& team) built it. :-) The ODT tooling is arguably the single biggest enabler of FLEXCUBE's phenomenal scale—creating an ecosystem that still thrives today. It allowed us to scale to an unprecedented number of customers, extend the product with agility, and implement projects in parallel. It paved the way for the large ecosystem of our Consulting teams and numerous partners who built on this foundation. Great technology often feels like magic. It just works. When RAD (ODT) first emerged, it was pure magic—how with just a basic data model and some simple configuration, it could create or extend (for modifications) a complete front-to-back module. From CRUD services to business logic to the user interface—everything came together auto-magically aligned to FLEXCUBE’s fabric. This magic was thanks to Radha and his team. At the root lay Radha's vision, an unparalleled end-to-end thought process. He wasn’t limited to technical details alone but truly understood the behavior of developers—how programmers approached FLEXCUBE development. He built non-invasive change management and seamless version control into the very fabric of the tool. This holistic approach is what made RAD/ODT the phenom it is today. Truly, there is no parallel for this kind of work in FLEXCUBE’s rich history—and Radha was at the heart of it all. After leaving i-flex/Oracle Financial Services, Radha went on to do more exciting work. While I don’t know all the details, I’m certain he continued to leave his quiet, indelible mark on every thing he touched. Always working his magic—unnoticed but unforgettable. More than anything, Radha will be remembered for his unassuming nature and constant smile. Radha, I will cherish the many conversations we had over the years as you built this phenomenon. Your loss feels unreal—gone too soon, my friend. I will also carry a personal regret: when you reached out for help just a few months ago, I couldn’t help. That will stay with me. Go well, my friend. I suppose God had need of a little extraordinary magic—and He called upon you to build it.

275

60 Comments

🔧 Exciting News! I just completed the Liquibase and the CI/CD Process course!  🎓 This course is perfect for anyone eager to integrate Liquibase seamlessly into their CI/CD pipelines. I delved into core DevOps and CI/CD concepts, learned how Liquibase fits into the process and explored automation tools to streamline database changes. Plus, I mastered source control management best practices and database testing for the Liquibase build environment. If you're diving into CI/CD and #DevOps, don't miss out on how Liquibase can level up your database automation! 🚀 #Liquibase #DatabaseAutomation #TechSkills

7

There are lots of latest technologies used in project or proposed in projects just because they are popular not because they are needed. Simple example is Kafka , it is highly popular now a days so it is proposed so often. But there is Active mq as well rabbit mq who also can be proposed for doing similar kind of work. It is important to understand what kind of application demands this technology that is very important before proposing it . If we can achieve same work with less complicated technology then it actually reduces time as well efforts

10

Digging into JHipster 8.3.0 and found a bug in their recent adoption of the Spring Boot Parent (SBP). The tracking to SBP reduces maintenance friction in the JHipster Dependencies BOM and tracks closer to Spring’s own security patches. Yet, the SBP brings in Spring’s own configuration of Maven’s Resources Plugin that JHipster was largely managing for legacy reasons. The merged effective POM resulted in JHipster’s resources plugin filters failing to execute as expected when activing Spring Profiles via Maven’s own -P/--activate-profiles switch. Compounding the problem, and hiding the issue, was using JHipster’s service discovery support with spring-cloud-starter-bootstrap. The bootstrap configuration was ignored by SBP’s filtering and JHipster “worked.” What I love about OSS is being able to identify the problem and submit the fix and at the same time pull JHipster closer to Spring Boot. The PR was approved and merged this morning! 🎉 #jhipster #opensource #springboot

25

2 Comments