Nirasha Jayalath@nirasha_jayalath26

Robot Abstraction allows plant operators to convey their intent for plant objectives to multiple robots in a plant in natural language and for Gemini LLMs using Google ADK to translate into commands that a supervisor agent communicates with different real robots in simulators and the real world. Robot Abstraction enables - Multiple robots from different manufacturers with similar and different capabilities be used. Enables new robot manufacturers to get adoptions - Research institutions to envisage complex workflows that require orchestration to be built just by conveying human intent in natural language - Plant operators to test what if scenarios in simulation environments such as IsaacSIM, Gazebo and webots for their specific factory configurations before making buying decisions. Robot Abstraction is the interface between human intent and robot action through the robot abstraction framework. Robots are physical systems. Mistakes have real consequences. LLM outputs can be unsafe or incorrect. APIs are inconsistent across platforms. Simulations don't perfectly match hardware. Most teams end up rebuilding safety systems from scratch. Robot Abstraction addresses these issues in a simple framework that constrains to typed capabilities such as navigate, stop, inspect - validates actions through policies before execution and guarantees safety and correctness through guard rails. Robot Abstraction starts with simulation first developers and pick and place robots before we move to connecting with physical robots and to humanoid and drones

RescuOrch integrates Webots R2025a simulation with DJI Mavic 2 Pro drones, TIAGo++ ground robots and the Gemini 3 Flash LLM to orchestrate multi‑agent rescue operations. Its current demo tackles a kitchen fire, but the same platform can simulate active‑shooter drills, FEMA‑style flood recovery, massive fires in dense areas like Mumbai’s Dharavi, pipeline or oil‑rig accidents and high‑magnitude earthquakes. Gemini produces real‑time plans, so drones scout hazards, ground robots execute tasks and the plan adjusts as conditions change. These simulations address real gaps: U.S. fire‑response times average 6–8 minutes in cities and exceed 10 minutes in rural areas; Mumbai’s fire brigade reports response times of ~10 minutes in the city and 20 minutes in suburbs. Globally, over 180,000 people die from burn injuries each year and 86,473 people died in disasters in 2023. Last year 89 U.S. firefighters died on duty, underscoring the dangers responders face. After earthquakes, survival drops from about 90 % in the first day to 5–10 % after 72 hours, so rapid coordination saves lives. By running “what‑if” scenarios—larger or multi‑room fires, stronger earthquakes or multiple hazards—RescuOrch helps agencies test strategies and decide if additional drones or rugged robots would improve outcomes. With 27,000 U.S. fire departments and 23 oil refineries in India, there is a broad user base for physics‑based simulation training. In short, RescuOrch offers a versatile AI‑driven testbed to help responders plan, train and procure the right equipment for complex emergencies.

RescuOrch integrates Webots R2025a simulation with DJI Mavic 2 Pro drones, TIAGo++ ground robots and the Gemini 3 Flash LLM to orchestrate multi‑agent rescue operations. Its current demo tackles a kitchen fire, but the same platform can simulate active‑shooter drills, FEMA‑style flood recovery, massive fires in dense areas like Mumbai’s Dharavi, pipeline or oil‑rig accidents and high‑magnitude earthquakes. Gemini produces real‑time plans, so drones scout hazards, ground robots execute tasks and the plan adjusts as conditions change. These simulations address real gaps: U.S. fire‑response times average 6–8 minutes in cities and exceed 10 minutes in rural areas; Mumbai’s fire brigade reports response times of ~10 minutes in the city and 20 minutes in suburbs. Globally, over 180,000 people die from burn injuries each year and 86,473 people died in disasters in 2023. Last year 89 U.S. firefighters died on duty, underscoring the dangers responders face. After earthquakes, survival drops from about 90 % in the first day to 5–10 % after 72 hours, so rapid coordination saves lives. By running “what‑if” scenarios—larger or multi‑room fires, stronger earthquakes or multiple hazards—RescuOrch helps agencies test strategies and decide if additional drones or rugged robots would improve outcomes. With 27,000 U.S. fire departments and 23 oil refineries in India, there is a broad user base for physics‑based simulation training. In short, RescuOrch offers a versatile AI‑driven testbed to help responders plan, train and procure the right equipment for complex emergencies.

GraphRisk: AI-Powered Partner & Affiliate Fraud Detection (Real-time + Explainable) Partner and affiliate fraud is fundamentally different from typical client fraud. It’s not about stolen cards or fake IDs—it’s about exploiting business relationships, attribution chains, and commission structures. Fraudulent partners often understand platform rules, distribute suspicious activity across many referred accounts, and stay just under detection thresholds. The result is delayed detection, high investigation burden, and months of commission leakage before patterns become undeniable. GraphGuard addresses this by treating partner fraud as a relational, network-based problem rather than a set of isolated transactions. The system builds a dynamic partner–sub-affiliate–client graph and continuously analyzes patterns across accounts, affiliates, and time windows to surface coordinated behavior that manual reviews cannot scale to find.

QQuest is a Proof-of-Quest platform that turns community participation in the Qubic ecosystem into verifiable on-chain achievements. Projects use QQuest to design campaigns made of quests such as “learn”, “code” and “participate” tasks. Users connect their wallet and social accounts, complete actions on Discord, Telegram, X or on-chain, and QQuest verifies them automatically using EasyConnect workflows and tools like Zapier/Make. A C++ smart contract on Qubic escrows QX tokens or NFT badges and releases rewards only when a quest is truly completed, preventing duplicate claims and basic fraud. Gamification features like XP, streaks, badges and leaderboards keep users engaged, while an analytics dashboard helps projects understand which quests drive activation and retention. QQuest becomes a reusable engagement layer for any Qubic dApp launch or community, combining AI-generated quests, social automations and secure on-chain rewards.

CodeMentor AI is an interactive coding platform powered by Codestral from Mistral AI, GPT o1 Reasoning focusing on C++ language learning (Further enhanced for any application ). It offers step-by-step coding tutorials with real-time feedback, adaptive project-based learning, and detailed performance analytics to track progress. The platform features a code playground for hands-on practice, code reviews for quality assurance, and the option for students to receive real-time help. Learners earn marks for their answers and benefit from personalized study plans to enhance their skills and understanding effectively.

Introduction In an era where infectious diseases can rapidly escalate into global pandemics, the ability to make swift, data-driven policy decisions is more critical than ever. We present a groundbreaking multi-agent epidemic simulation tool designed to empower policymakers with actionable insights, enabling them to navigate the complexities of epidemic management effectively. The Problem Policymakers face significant challenges during epidemics: Complex Dynamics: Epidemics involve intricate interactions among individuals, communities, and policies, making outcomes difficult to predict. Uncertain Outcomes: Without robust predictive tools, policies may lead to unintended consequences, exacerbating the crisis. Data Overload: The abundance of fragmented data sources complicates the decision-making process. These issues can result in delayed responses, increased transmission rates, and greater societal and economic disruption. Our Solution Our simulation tool addresses these challenges by: Multi-Agent Modeling: Simulating individual behaviors and interactions to capture the nuanced spread of diseases. Policy Impact Analysis: Allowing users to model and compare the effects of various intervention strategies, such as lockdowns, vaccination campaigns, and travel restrictions. Real-Time Insights: Providing immediate feedback on potential outcomes, helping to adjust policies proactively. User-Friendly Interface: Designed for policymakers without requiring technical expertise in modeling or epidemiology.

CodeMentor AI is an interactive coding platform powered by Codestral from Mistral AI, focusing on C++ language learning (Further enhanced for any application ). It offers step-by-step coding tutorials with real-time feedback, adaptive project-based learning, and detailed performance analytics to track progress. The platform features a code playground for hands-on practice, code reviews for quality assurance, and the option for students to receive real-time help. Learners earn marks for their answers and benefit from personalized study plans to enhance their skills and understanding effectively.

CodeMentor AI is an interactive coding platform powered by Codestral from Mistral AI, focusing on C++ language learning (Further enhanced for any application ). It offers step-by-step coding tutorials with real-time feedback, adaptive project-based learning, and detailed performance analytics to track progress. The platform features a code playground for hands-on practice, code reviews for quality assurance, and the option for students to receive real-time help. Learners earn marks for their answers and benefit from personalized study plans to enhance their skills and understanding effectively.

📅 May 11–19, 2026 • May 11 – 19: Online Build Phase. • May 18 (Hybrid Build Day – Online & Onsite) • May 19 (Demos & Awards – Online & Onsite) 🌍 Hybrid event. Join online from anywhere or build onsite at the venue. 🏆 Winners will be announced live on stage at the AI & Big Data Expo North America on May 19 📍 On-site Venue (May 18–19): San Jose McEnery Convention Center, CA, USA 🤝 Join solo or with a team. 🎟 FREE AI & Big Data Expo North America ticket included for participants that are able to join the on-site phase. 🏆Compete for a $10,000 prize pool 🧑‍💻 Apply now to design, prototype, and showcase enterprise-ready innovation.

Grab your ideas, spin up compute, and start building what’s next! Start with $100 in AMD Developer Cloud credits and begin building right away. Get access to hands-on tutorials, workshops, and curated learning resources to help you build faster and go further. To unlock credits and join the hackathon, sign up for the AMD AI Developer Program. 🤝 Go solo or team up — ship something real. 💰 $21,500+ total prize pool & 1 × AMD Radeon™ AI PRO R9700 GPU up for grabs. 📍 On-site Venue (May 9-10): MindsDB SF AI Collective 3154 17th St, San Francisco, California, USA 📍On-site participation is by invitation only. Travel and accommodation expenses will not be covered. 🧑‍💻 Developers, builders, and innovators - bring your boldest ideas and let’s turn them into real AI systems on AMD.

Teams work together with multiple AI agents in shared workspaces to design, build, and launch complete AI products — from initial idea to market-ready app. 📅 February 23– March 1, 2026 ⏱️ 6 days to build and ship an end-to-end agentic AI product 🤝 Work together in shared AI workspaces where your team and AI agents collaborate in real time 🔧 Build with multiple AI agents that handle development, design, testing, and documentation together 🎯 Create everything in one place — from ideation to product to marketing materials, all by prompting and observing 🚀 Focus on real-world products that teams could actually ship and use 🏆 $3,500 in prizes + premium 3-month Complete.dev access for top teams 🧑‍💻 Join your team workspace and start building together