I am taking some time to explore different problem spaces. I am especially interested in artificial intelligence, neurotechnology, and other deep tech domains that are on the precipice of technological advancement. That being said, I am a man of ideas, so below I have shared one main idea, followed by some shorter ideas, and then a link to my internal repository of ideas. The cool thing is that I am currently working on building all of these.

  • A better search. The vast world of data we now find ourselves in has exhausted the archaic approach we had once taken towards search. Oftentimes, Google searches are riddled with poor quality content that played towards SEO and maybe even paid some money to be put on your screen. After diving into these links, it is your responsibility to find and understand any relevant information from the article. Instead, we should be able to ask questions about pre-existing information tailored to exactly what we want. The new paradigm of search will include internal search engines for documents owned by a single company.

    With the use of semantic search and large-language models (generative AI), I am creating a way to upload, parse, and train on any document such that users can ask questions about any information that might be written, and receive answers in a humane way. The broader vision here is to be able to eventually combine these smaller internal search engines to be able to ask questions about any information in the world and receive a response directly tailored to how you like it.

    Since the idea is quite broad, there is an easy niche to target first. Currently, lots of time is wasted in companies by asking questions that are already outlined in the employee handbook. Think back to a time that you asked your HR manager what exactly your insurance covered, or if you got paid overtime, or how your equity vests. Since all this information is already clearly documented, I would like to ask companies to submit their internal documentation to create an AI employee concierge for any questions. Another vertical that we could attempt to target with this is techincal documentation.

  • Using traditional NLP and large language models to automatically test students about some knowledge domain and poke holes their understanding. Eventually, this will be a faster and more accurate way to test the understanding of a concept that a student has without taking a toll on teaching resources. Later on this will transform into a machine that queries students about the exact right topic to accelerate the progress of learning.

  • Creating a social media that doesn’t require the internet. This will be done at first by using SMS, and then transitioned to a peer-to-peer Bluetooth network that allows for users to access information without internet. I am currently building this out (and may even ship it before you read this) for the people in Iran facing internet outages as a result of political unrest.

  • Using cameras, eye tracking, and brain signals to detect the emotion of a user. This emotional data can be used in creating an emotional passport with a lot of data about when a user has heightened emotions (while playing games, interacting with one another, etc). This data could be very useful for designers of any product. It could also be used to create a core memory recorder that automatically records the activity of the user (in real life, a video game, or metaversal experience) based on when they have a heightened emotion

  • Creating personal assistants using fine-tuned large language models. I am currently building this in the form of a bot that automatically replies to texts for me based off of all of the writing I have ever recorded.

  • increasing information transfer rate of brain-computer interfaces with large language models. This is currently a research project for me.

Please see more ideas at ideas