Assistive technology & entrepreneurship
Drishti · Oculosense
Computer vision and hardware built around accessible interaction.
Assistive computer vision for people with visual impairments, connecting early navigation prototypes with the development of AI-powered devices at Oculosense.
Co-Founder & Chief Technology Officer · September 2021–August 2023
Prototype: 2020–2021 · CTO role: 2021–2023
Prototype-to-product journey
12 FPS refers to my early prototype; deployments describe Oculosense’s current product, including later development.
Drishti user stories.
Watch on YouTube01 / CONTEXT & CONTRIBUTION
Drishti: Assistive Vision at Oculosense
The early Drishti navigation prototype explored real-time obstacle detection and a non-invasive muscle-actuation mechanism for pedestrians with visual impairments.
My subsequent work at Oculosense involved co-founding the company and leading technical development of assistive devices that combine computer vision with hardware.
My role in the work
- I designed an assistive navigation prototype and implemented TensorFlow Lite/YOLOv4 object detection at 12 frames per second.
- I co-founded Oculosense and led technical development of AI-powered assistive devices as CTO from 2021 to 2023.
- My work connected computer vision, hardware integration, and product development for accessibility.
02 / METHOD & SYSTEM
From perception to assistive hardware.
01Prototype around a navigation task
The early project combined obstacle detection with a navigation-control mechanism, turning visual perception into an assistive interaction.
02Integrate perception with hardware
I worked across the vision pipeline and device integration, accounting for the constraints of wearable and embedded hardware.
03Design accessible interaction
Drishti Vision now combines text reading, scene/object recognition, familiar-face recognition, and audio responses controlled from the glasses.
03 / RESULTS
Prototype performance and product evolution.
- My early navigation prototype ran object detection at 12 FPS.
- Since my 2021–2023 CTO tenure, the product has continued to evolve. Oculosense now lists more than 1,000 Drishti Vision deployments across India.
- Oculosense lists the current Drishti Vision at 49 g, with support for more than 100 languages.
What I take from this work
Drishti represents my experience taking computer vision into assistive hardware and product development through entrepreneurship.
04 / DESIGN CHOICES
Building for accessible interaction.
Run perception on the prototype hardware
I implemented real-time detection with TensorFlow Lite and YOLOv4, reaching 12 FPS in the early navigation prototype. That made model execution and hardware integration part of the same engineering problem.
Connect detection to a navigation response
The original design paired obstacle detection with a non-invasive muscle-actuation mechanism. The technical task extended from recognizing an obstacle to connecting that detection to an assistive response.
Follow the device into use
The Drishti user video shows glasses, earphones, and a connected handheld unit being used together. It gives a view of the complete interaction, while the hardware-evolution image traces the product beyond my 2021–2023 CTO role.