I am really lucky to have to been selected for an elite Nanodegree program about self driving cars conducted by
Udacity. More than 12,000 applicants have applied for this program and only 250 students were selected. The course runs
for 9 months.I am really excited and thrilled about this opportunity :)
I will work on deep learning, computer vision and lot other subjects. Here is a link to detailed curriculum :
A new powerful method to predict major earthquakes
I single-authored a paper on Deep Learning, and it has been published in arXiv. The paper concerns the application of Deep Learning in earthquake prediction. Earthquakes are classified as foreshocks, mainshocks and aftershocks. Most of the mainshocks are preceded by foreshocks making these earthquakes good precursors for mainshocks. So far, there is no method of identifying foreshocks, making it impossible to predict the occurrence of the more devastating earthquakes. I trained the network to identify and classify earthquakes as foreshocks, mainshocks or aftershocks instantaneously, i.e. as the events occur. The algorithm has successfully identified foreshocks with just seismic waveform as input. This algorithm, when applied in real-time, would greatly help in predicting imminent earthquakes.
link to paper
Low cost solution to identify refractive errors
One of the major cause for blindness is due to refractive errors. All the available devices to detect them are pretty expensive and bulk. A team at Srujana Innovation center, which has a tie up with MIT Media Lab came up with a
solution called "JULIA"-Just use logic for Imaging Applications. The current version had shortcomings in measuring power, some ergonomics and necessity to focus at an image 3m away. I got a chance to work on the latest iteration during the "Engineering the eye Workshop" conducted at LV Prasad eye institute from 3-9th July, 2016. I worked on the optical design and together as a team we redesigned Julia from scratch to a unibody stand alone design. Using big data approach we incoporated human errors making Julia readings reliable. The best
part is that, the total cost per device can be scaled to Rs.20/-
New channel to learn Deep Learning with TensorFlow
I have launched a new YouTube Channel for learning and understanding deep learning with TensorFlow. The goal is to discuss deeo learning algorithms and their code implementation from scratch,
to motivate everyone to learn, code and build applications based on deep learning.
Watch the pilot episode here:
My first half marathon
I completed my first half marathon at MEPZ, Chennai link with a theme to make MEPZ green and clean. About 300 people participated in the event. My timing was 2hrs 43mins. My next target is to
increase the pace and aim for full marathon :)
I've Joined Ligt Matter Interaction at OIST,Japan as a Technical Staff. I would be working in different fabrication techniques and studying optomechanical effects.
Really excited about this !!
New Delhi, 07/06/2015
I successfully completed duathlon, cycling for 27kms followed by 10kms running. This is the very first time, I extended myself and
biked and ran back to back. Happy and excited to have completed a duathlon.
IIT Delhi Research
I completed a three month research internship at the Photonics Research Lab, Physics department at IIT Delhi.
New Delhi, 10/05/2015
I ran 5K run on Mother's day. This was my first 5k run and hope to run longer The theme of the run was to make cities safer for woman