Co-founding a startup as an undergraduate student

How it all started

In 2020, during my first year as a computer science undergraduate student in India, in the middle of the coronavirus pandemic, I was desperate to find productivity again. Encouraged by my father, I began learning deep learning concepts and computer vision through Andrew Ng's Coursera lectures and a Udemy crash course. I was pretty consistent and got some decent hands-on experience using Tensorflow and image processing libraries like OpenCV. Around that time, my father was interested in creating an AI-driven solution for simplifying karyotyping, a commonly done cytogenetics test. I decided to put my new programming skills to test and developed a basic prototype of a chromosome image classification model by training SqueezeNet, a deep learning model. A genetics clinic responded positively to this idea and before I knew it, we were in the ideation stage of a MedTech startup. 

It was during Summer of 2022 when I, along with a small group of friends, progressed well in creating an initial working product that accepts a metaphase image, performs chromosome segmentation, feeds the segmented images to our computer vision model, and produces a karyotype.  I trained state-of-the-art deep learning models for classification and handled segmentation of complex metaphase images. My novel segmentation pipeline was a combination of seed-based watershed segmentation and autoencoder image reconstruction to separate overlapping and clustered chromosomes. 

As the startup grew, we were selected as one of the top 60 startups globally to be incubated at Berkeley SkyDeck, which is UC Berkeley's startup accelerator program. I took a temporary break from classes and assignments (yay) and we visited Berkeley for a month to participate in the program. I had the opportunity to interact with various early-stage startups from across the world, hear their stories and business strategies, and get mentored by faculty at UC Berkeley. Overall, this was an incredible experience, and it showed me what the startup ecosystem looked like and how to navigate this long but rewarding journey to become a successful startup.

Since then, CARAIO Technologies has flourished in many ways. As technology improved, we adopted it into our product. We continue to interact with numerous cytogenetics labs and hospitals, receive feedback, and improve our product. Being involved in building an AI MedTech startup from day one has been both thrilling and intellectually stimulating. It has helped me develop skills in computer vision, product development, and several other areas.

Outcomes

• A. Pallavoor, P. Anagani, S. Tambarahalli, S. Pallavoor. Chromosome Segmentation Analysis Using Image Processing Techniques and Autoencoders. American Journal of Computer Science and Technology, 5(4), 210-216., 2022. (link to manuscript)

•  Selected Participant, UC Berkeley SkyDeck Pad-13 Batch 15 & 16 Incubation Program (2022-2023) 

•  A. Pallavoor and S. Pallavoor AI for Karyotyping: Automated Chromosome Analysis Using Deep Learning. Poster presented at BIOMEDevice Conference, Santa Clara (2022)

•  Developed a Cytogenetics Copilot chatbot with GPT-4, LangChain, and Pinecone as part of UC Berkeley Cal Hacks (2023)

• Participated at the G20-DIA Summit, Bangalore, India (2023)

• AI Keynote Speaker, iCARE-HS International Conference (2024)

• Invited Speaker, UC Berkeley ISPMA (2024): Challenges and Opportunities in Bringing AI to International Markets