Cognitive AI | LinkedIn (original) (raw)
Software Development
Berkeley, CA 114 followers
Smarter Tools for Stronger Minds
About us
Cognitive AI is a startup creating engaging and accessible AI tools to be used in clinical psychotherapy settings. We align AI on techniques derived from behaviorism and cognitive psychology to supplement behavior therapy. The Cognitive Distortion Journal (The CDJ), a flagship product of Cognitive AI, is a tool that identifies cognitive distortions in every day thoughts and helps to reframe them; a common assignment in cognitive behavioral therapy. Interactive tools can enhance the efficacy of psychotherapy. Such tools are capable of boosting assignment completion rates and offering immediate feedback when necessary, courtesy of AI-powered extensions of the therapeutic process.
Industry
Software Development
Company size
1 employee
Headquarters
Berkeley, CA
Type
Self-Employed
Founded
2023
Specialties
Software Development and Psychotherapy Tools
Locations
Employees at Cognitive AI
Updates
-
Founder & SWE | BS in Software Engineering
8mo Edited
One of the most enjoyable parts of building (generative) AI apps is the level of creative freedom you feel you have. The feeling of "the right way to do things" is pretty much absent because no one has really done this before, and that is quite liberating. In my application, The Cognitive Distortion Journal—which you can request beta access to at https://lnkd.in/g6RTc9UD —I'm designing a system to generate weekly reports on users' entries, allowing users to view any particular thought patterns they may have struggled with that week. The diagram I created details the flow of what such a request might look like (everything in the green box is happening on the backend). You can view my GitHub issue, https://lnkd.in/gmtN4s29, for more detailed information. The highlight is how the generative AI model interfaces with the backend through a custom, prompt engineered, GPT model, CdGpt (inherits from Assistant). What's mind blowing to me about this is that I'm designing an AI interface that has private route access to my API. My job in this process is architecting the flow. I just have to focus on engineering optimal channels, building the path between client and LLM. I don't have to design some fancy algorithm to build content atop user data, the LLM does it for me! Something interesting that occurred to me while writing this post is the idea of never having to touch user data. Provided the LLM does what you ask of it— which for the most part, it does, depending on how you instruct it—it can potentially maintain all server-side data. Now the question is, how willing are you to relinquish control to it? And how much of it should you? 🤔#GenerativeAI #LLM #GPT #TheCDJ #WebDevelopment #AIApps