This page features some of my side projects, including potential startup ideas, hackathons, and learning initiatives.
Previous Projects
Following is the running list of ideas on my to-do list. Some have the potential to evolve into full-fledged startups, while others are fun hackathon projects. If you would like to team up or contribute to any of these ideas, please feel free to reach out.
Robotics and Deep Tech
Bio-mimetic robotic hand: Despite significant advancements in robotics, achieving the same level of dexterity as a human hand remains a major challenge. This represents a bottleneck in adopting and applying robots for everyday use. However, we can address this challenge by developing a robotic hand that closely mimics the human hand in terms of its joints, degrees of freedom (DOF), and motion ranges. Read more
Dexterous, personal robot chef: Everyone deserves to enjoy delectable, soul-nourishing meals. Yet, not everyone possesses the skills, time, or resources to create gourmet dishes. Enter the ultimate solution: a robotic chef powered by AI and outfitted with a pair of dexterous arms mounted on a versatile mobile platform. The AI can be trained over millions of cooking videos on the internet along with the required skills to use everyday kitchen items. Read More.
Generative AI
Flute covers of songs using AI: I am always searching for the perfect jam that hits the sweet spot between a purely instrumental melody and a full-on lyrical banger. Flute covers like this seem the optimal choice, as they allow me to focus while working and provide an option to hum along to the beat if I want to. We can mass-produce them for all popular songs using AI foundation models such as MusicLM.
2-min video gist from long videos: I would love a concise, 2-minute summary of lengthy YouTube videos, complete with key highlights and visuals. A textual summary of the transcription falls short because it (1) requires reading rather than watching and (2) omits important visuals, such as graphics or actions displayed in the video. To generate such 2-min clips, we can use GPT-4, which processes image and text inputs, alongside VALL-E to retain the original speaker’s voice.
Software tools and apps
Interactive multimedia PDFs: PDFs are rooted in the 1990s, while modern communication increasingly relies on multimedia content (e.g., videos, graphs, animations) and interactive elements (e.g., tabs, forms, dropdown lists, sliders) to convey information on the web. By integrating the most advantageous features of PDFs, such as unalterable and offline accessibility, with the modern UI/UX of web-based platforms, we can create a new content-sharing format that aligns with the demands of today’s digital landscape. Read More
Knowledge graphs to holistically show one’s skill sets: Resumes and LinkedIn fall short in portraying an individual’s dynamic and diverse skill set. By utilizing knowledge graphs, we can effectively capture and present a comprehensive view of one’s evolving abilities over time. Individuals can generate these knowledge graphs by monitoring their online activities and ultimately unlocking new professional and personal opportunities previously unattainable. Read More