For the USAII Global AI hackathon, we built an AI Benefits Navigator for older citizens. The idea came from something my teammate (Mujib ur Rahman) and I both noticed in our own families. As more essential services move online, some of our older relatives have found it increasingly difficult to navigate applications, forms, eligibility requirements and deadlines. It made us think about how much harder this must be for seniors living alone. Our solution helps seniors understand what support programs they may qualify for, explains requirements in simple language, remembers important documents and deadlines, and even helps connect them with relevant support services when needed. One thing I'm particularly proud of is the demo video. Out of all the hackathons I've participated in so far, this is probably my favorite demo we've created. It feels like it tells the story behind the problem, not just the technology. Shoutout to United States Artificial Intelligence Institute for organising such a fun hackathon! #AIforGood #Hackathon #DemoVideo

Had fun last weekend on the Gemma- Google DeepMind X Cerebras hackathon. Due to super fast (almost instant) inference, I was forced to thing outside the box and instead of going into the normal "multi-agents" o "counsel of agents" I decided to go "possible futures". Shipped Future Shield. Before a safety-critical AI acts, it forks N futures, probes each with Gemma 4 at 1,500+ tok/s on Cerebras, kills the bad branches, and collapses the survivor to mainline. Pre-action safety, in real time. Hardest part showing the idea in the UI. This is a great working prototype and I will use it for decision making!

You only know what a hackathon is like once you've been to a hackathon. So we wanted to do something different. Instead of just a recap or the final demos, we wanted to capture what actually happens in between: turning an idea into a working prototype under time pressure. What building products looks like in the AI era. How teams collaborate. And maybe what happens when you mix a little sleep deprivation with a lot of caffeine and adrenaline. We followed three teams building: 🗣️ A voice-first ordering experience 🍔 A new way to discover and share food recommendations 🚨 An AI-powered Incident Copilot for on-call engineers A documentary about how people build under pressure. 🎬 Episode 2 of What Happens in 48 Hours? — Behind the Hack is out now. Watch the full episode here 👇 https://lnkd.in/egxx7CRC #LifeAtWolt

Ever tried to build your own PC? Or a drone? You have to learn what every spec actually means. Then pick from thousands of valid builds, without guarantee yours is the best one. I'm thrilled to have taken part in my first Kaggle hackathon: the 5-day Vibe Coding Capstone Project, with a wonderful team: - kebei jiang, - Ayush Morbar, - Palak Sharma, - Alex Ahern We built an advisor agent combining domain expertise and operations research: Getting a quality product that truly meets your expectations takes a lot of time, a solid understanding of the specs, and extensive research into every possible configuration. Even with tremendous effort, finding the best setup among all the options can be long and difficult. Our agentic system makes this task: - simpler since it allows the user to discuss in natural language with the agent and refine their requirements. - faster, because the agent is able to model the problem quickly. - optimal, because a solver explores the full space of valid builds under your constraints. It is composed of: - Concierge Agent: Interacts with the user and models the optimization problem with a self-evaluation loop - Solver Specialist: Prepares the input data, calls the solver as a tool and outputs the result. The agent returns a whole config in a table, along with the reasoning behind each component. (Link to the report in the comments). This project gave us a deeper understanding of what's at stake in AI agent development (security, cost, deployment), and of how important these systems will be in the years ahead. It was also a real teamwork experience: scheduling meetings across 3 time zones, with diverse profiles that require communicating effectively to move forward together. I sincerely want to thank them for their work, we learned a lot together, and it's an experience I would absolutely do again! And of course, thanks to Kaggle and Google for this comprehensive course on a topic of growing importance. #kaggle #agents #hackathon

This should be a great opportunity for UCSC students interested in gaining experience with Google DeepMind's open Gemma models, check it out!

We're running our first joint hackathon with Trigger.dev and I couldn't be more excited about this one. 7 days, fully virtual, EUR 10,000 in prizes , and a challenge that gets right at the core of what makes AI agents usable: speed on the data side, reliability on the execution side. If you've ever built an agent that worked great in a demo and then fell apart the moment a task ran long or a query got slow, this is for you. Registrations close July 16. Go build something. 🔗 luma.com/clickh-uko4

Have you ever hesitated to post your work online because it didn’t feel “ready”? This post is for you. For the Bubble #PromptAndCircumstance hackathon, I created half.done, a small internet corner for unfinished ideas, abandoned projects, experiments, and things we never felt were “ready.” Contra Most of us wait for perfection before sharing our work. We keep projects hidden because they feel incomplete. half.done is a place to post the things you loved, learned from, and worked on, without the pressure of a polished ending. I’d love to see the ideas, side projects, and half-finished things you’ve never shared. try it here https://lnkd.in/gmrWB-Ci What surprised me most was how it came together: it started as a single Bubble AI prompt. From there, I kept reshaping it step by step, turning a basic generated structure into something usable, expressive, and actually fun to use.

Excited to share that I participated in the Semicolon 24-Hour Hackathon! 🚀 Over 24 hours, my team and I worked on solving a real-world customer retention challenge for dummy client website called as ShopTaal, an online music instrument store. Our solution focused on building an “Al powered Customer Risk Agent” that: Continuously ingests customer activity such as orders, logins, ratings, cart activity, and support tickets. hl Uses multiple behavioral signals to predict churn risk (High / Medium / Low). e Automatically triggers personalized interventions like WhatsApp reminders, discount vouchers, loyalty rewards, or support callbacks. Learns from the outcome of every intervention to improve future customer re-engagement. The hackathon was an incredible opportunity to think beyond coding-understanding business problems, designing Al-driven workflows, and building practical solutions under a tight 24-hour deadline. Grateful to the organizers for hosting such an ens jing event and to my teammates for the collaboration, creativity, and late-night debugging sessions. Looking forward to participating in more hackathons and continuing to build impactul Al solutions! #Hackathon #Semicolon #AI #MachineLearning #CustomerRetention #AgenticAl #Innovation

How many ideas can your team come up with in just one hour? We decided to find out. During our internal hackathon, we gave every team the same challenge: come up with an AI tool or feature that could eliminate repetitive tasks and make everyday work a little easier. In just one hour, people: - formed teams of 2–3 - brainstormed AI solutions - discussed which ideas could have the biggest impact - pitched their concepts to the rest of the team Some ideas were ambitious, some were fun, and some were surprisingly practical. But they all started with the same question: What could we automate to make our work and life better? We've rounded up some of our favorite concepts in this carousel. Swipe through to see what our teams came up with 👉

WisdomAI engineering team is one of the most customer focused teams. The foundational framework enables us to rapidly build and deploy features across a large number of users.