Jerrold Vincent’s Post

Great panel discussion on AI in Nuclear at #NAYGN 2024 conference! Some of the key takeaways: - AI is already here in nuclear, being used by many utilities for lots of different use cases (CAP and MRFF screening, core reload analysis, documentation generation). - It's okay to use AI, even for regulatory uses (e.g. license renewals and applications) if there is a human that signs off on the work. GenAI is another tool in your toolbox. - "Trust but verify", the existing culture in nuclear is well suited for taking advantage of AI while avoiding the pitfalls. - NRC is planning to embrace AI technology over the coming years, want to enable the use of AI while being safe and responsible. - Defense in depth and additional verification processes are a pathway towards fully autonomous AI with human verification, but there is still a way to go both technically and from a regulatory perspective. Thanks Bradley Fox, Luis Betancourt, P.E., Michael Brasel, Raeven Whitfield

Nuclearn is one of 4 finalists for the 2024 AZ Inno Fire Awards in software! Thank you to whoever nominated us, I'm honored that we have been selected for such an exclusive group. Exciting! https://lnkd.in/grVjsVTV

I'm at the end of my 2nd full year at Nuclearn, and I've been reflecting on 2023 and felt like I'd share some of my thoughts. This was a year of big change and a lot of growth. We doubled our customer count, doubled the number of products we offer, made significant advancements applying fine-tuned large language models to Nuclear-specific domains, and so many more things that I don't have the space to enumerate here. We started the year with 1 employee besides Bradley Fox and I, and now our headcount is over 10 and growing. And through all of that... I've learned a few things. In no particular order: - Making decisions quickly and having a bias for action is really really important. In previous roles working for a utility, most decisions were hard to back out of and often took a long time to make (often for good reasons, e.g. longer planning cycles, emphasis on capital projects, etc). That isn't really true of an AI software startup, and delaying decisions is the death of agility. Most things we do are "two-way doors", and doing something >> doing nothing. The resulting snowball effect on momentum you get from more quickly iterating through your decide - act - observe loop is far more powerful than I ever would've thought. - Hire the right people, give them the explicit goals, and explain the reasoning behind those goals. When you give the right people challenging (but realistic) goals, they put in the extra effort to achieve those goals. When they understand the "why", they are able to prioritize and adjust on their own instead of needing explicit direction. And then when they meet those goals, they get much more fulfillment. I've seen this work so well in the last year, but still don't take this approach as often as I should. This is something I'm really going to focus on this year. - Unsurprisingly, startup life can be... chaotic. Lot's of hours, last minute trips, unplanned fires that need to be put out. It's hard to express how thankful I am of my family and especially my wife for all of the support. Managing a startup while raising 2 young children wouldn't be possible without you all. - 2 years in and I still love it. The challenges are always different, and the excitement rarely wanes. Maybe having great customers (you know who you are!) and an awesome team help a little with that. 2024 is going to be great. We have made some recent AI advancements that I think are going to "wow" a lot of people, and make a huge positive impact for our customers. And we have a strong core team that will be able to execute on the opportunities. 2024 here we come!

This is great news! Palisades was originally planned for decommission because of economics - the fact that Holtec was able to sign a Power Purchase Agreement (PPA) for up to 2/3rds of the output of Palisades is direct evidence that Nuclear can compete as part of the domestic energy mix. https://lnkd.in/g_AmXZrk

✨ Exciting news! Nuclearn is seeking a Full Stack Engineer (React, Python) to join our team. We're working to revolutionize the nuclear and utility industries with innovative AI solutions. Passionate about technology and eager to make a difference? Apply now and let's shape the future together! 💻 #hiring #fullstackengineer #AI #jobopportunity

The thing I'm most excited about is that we've made an AI product that actually solves a real problem and is easy to use. When we started Nuclearn 2 years ago we began with a Machine Learning model that could take Corrective Action Program (CAP) text and predict related trend codes (Performance Objectives & Criteria). We thought that would be our product - Nuclear specific ML models we could deploy on premise for utilities to integrate with their processes. Our (rather bland) name for this product reflected just that: "Automated Trend Coder". We found out very quickly that was barely the first step to delivering a value added solution. Customers now had all of their CAP data organized in neat buckets aligning to key performance areas... but what were they to do with that information? We helped our first couple customers analyze that data in spreadsheets and pivot tables, and immediately realized that our product didn't solve their actual problem. The problem our customers needed solved was not "bucket our CAP data using AI", it was "help us identify and fix our performance issues". To do that, we had to build a lot of new product! We started with performing the necessary data aggregations and calculations via a built-in Analytics module (v1.4). Next, we built a custom "Cluster Visualization" that allows users to quickly identify performance issues visually (v1.5). Finally, we built a reporting module that outputs performance issues, supporting data and visualizations into a formatted, editable document, allowing users to share their findings, begin investigated issues and generating action plans to resolve (v1.6). We never envisioned taking the product in this direction, but this is what our customers needed. We listened to them, iterated our solution, and have been able to deliver an AI-based product that actually meets their needs. We now call this product "Automated Assessment Readiness and CAP Monitoring", and are very proud to have this product deployed at 27 reactors across North America. We've got a lot of new product ideas and AI capabilities we will be trying out this year, and I'm looking forward to continue making sure they solve real, meaningful problems for our customers!