👋 Hello there! I’m Prajwal.

I’m a 17-year-old based in Vancouver and obsessed with building projects that apply large language models to solve problems, big or small. There’s lots of LLM technology out there, and I want to bridge the gap between the tech and the real world. Right now, I am implementing LLMs to help visually impaired people.

LinkedIn Substack Medium X

Demonstrating the first prototype at the Vancouver AI Community Meetup in May

It has been almost 11 months since starting this project.

In January, I wanted to combine my knowledge of building LLM applications and using my Raspberry Pi hardware which I had. LLMs had gained the ability to understand images, respond with low latency, and many prompting techniques had been established.

Looking at this, I saw the opportunity of building an AI voice assistant with a camera, which could understand its surroundings and answer questions. What’s different with this approach is that I found an idea, product, or solution before thinking about the intended audience.

However, with the ability to see, the AI could assist people who have trouble seeing or are legally blind.

Bridging the gap between visually impaired people and society

• In the U.S., approximately 7 million people experience vision loss, with about 1 million classified as blind.
• By 2050, they are projected to reach 11.4 million, and it is estimated that between 27% and 43% of uncorrectable visual impairment will be attributable to myopia.
• In Canada, 2.6% of children aged 6 to 19 have visual impairments that cannot be improved to a score of 20/40 or better even with corrective lenses.

<aside> 👀

64.2% of visually impaired elderly individuals in the U.S. used informal caregivers, which is almost three times higher than the rate for visually unimpaired elderly (22.88%) [source]

</aside>

While many visually impaired individuals lead independent lives by many techniques such as using mobility devices and understanding the environment without vision, there are a significant number of individuals who receive assistance from assistive technology and caretakers.

JARVIS can function as an AI caretaker, assisting visually impaired people when caretakers are not available, and helping the caretakers better assist them.

In the first half of this year, I focused on building a prototype.

My first prototype, a Raspberry Pi on a hat with a camera, speaker, microphone, and button, running Python code.

I received the Ingenious+ Regional Award grant from the Rideau Hall Foundation and the BC Lieutenant Governor.

I have demonstrated JARVIS multiple times at Vancouver AI Community Meetups, increasing my presence.

Building this project:

1. Increased my technical knowledge about how LLMs process language and how to implement them as assistants and agents,
2. Has given me motivation and a grant to continue building this project and make it a reality
3. Established myself as a young builder in the Vancouver AI community
4. Enabled opportunities such as the Boreal Express and Hack the North