Thank You For Adjusting Innovators Judges
If you are reading this, it means I am doing a live pitch at another competition. I will be back within less then 10 minutes.
Please watch the 3 minute pitch video that I prepared. In case I am not back to answer your questions, there are FAQs below and the pitchdeck attached so you can scroll through it.
Comprehensive Slidedeck
All demo videos in the slidedeck pdf are attached below (scroll down)
Slide 12 Basic Interaction
The user's instruction is executed successfully on the first iteration. The user asks to find some halloween candy. Instead of simply searching for halloween candy in the search bar, the LLM effectively navigates to the Halloween/Halloween Candy page, showing it can effectively make decisions based on the page content.
Slide 16 Error Handling
If the LLM does not accurately perform the action on the first iteration, the support LLM kicks in to resolve the errors. In this example, the LLM unsuccessfully tries to click 'shopping cart' and the Support LLM properly corrects the action to 'add to cart' based on the available page elements.
Frequently asked questions
What industries beyond assistive technology can DexAssist disrupt?
DexAssist’s agentic framework is highly adaptable and applicable beyond accessibility. In healthcare, it can function as a virtual nurse—guiding patients through intake forms, scheduling, or medication reminders. In software development, it can act as a coding assistant that builds UI components or debugs code autonomously. More broadly, DexAssist can streamline workflows in education, customer support, enterprise productivity, and legal tech—anywhere users interact with complex digital systems.
How does DexAssist adapt to the needs of each unique user?
DexAssist uses personalized memory structures that evolve with user interaction. The agent learns from user behavior—such as task preferences, timing, and navigation habits—to create tailored support for common actions. Over time, it can anticipate user needs and reduce friction in digital tasks. This adaptive behavior is at the core of our agentic design philosophy.
What platforms does DexAssist support?
Our current version is available as a Chrome web extension, optimized for quick deployment and broad accessibility. We're actively building a backend to support cross-platform expansion, including mobile and desktop environments. By late 2025, we plan to introduce DexAssist for native applications, starting with Windows. Long-term, we aim to make it OS-agnostic and integration-ready.
What is the long-term vision of DexAssist?
Our long-term vision is to make computer navigation fully accessible for individuals with fine motor impairments—ensuring they can complete complex digital tasks independently. Beyond accessibility, we aim to increase business productivity by developing intelligent agentic agents that can automate workflows, reduce manual input, and assist across a range of enterprise applications.
What memory systems are used in DexAssist's agentic agents?
DexAssist agents are built on a hybrid memory framework that incorporates episodic, contextual, and semantic memory structures. These systems allow agents to recall specific user actions, understand intent across sessions, and generalize knowledge to similar tasks. We’re actively testing which configurations produce the most reliable and human-like behavior. Our memory-first approach is what differentiates DexAssist from traditional task bots or static assistants.
Why does DexAssist publish its technical research to the public?
First and foremost, we do not publish all of our novel or proprietary work—only the components where we seek feedback from industry leaders and the research community. Presenting at conferences has already helped increase DexAssist’s visibility and credibility in both academic and professional circles. This exposure directly led to key partnerships with the University of Illinois Urbana-Champaign, the University of Michigan, Slalom Build, and other collaborators. Publishing select work is part of our strategy to build trust, foster innovation, and stay at the forefront of AI and accessibility.