I’m a Senior Software Engineer at Meta Platforms Inc. (formerly Facebook). I am part of Cross-Meta Support organization (previously Central Customer Support). I am part of the team responsible for providing support to Facebook users who have lost access to their accounts. I am involved in leveraging Artificial Intelligence [AI] and Machine Learning [ML] to scale the product globally.

Prior to Meta, I was a Software Developer Engineer at Amazon Web Services(AWS). Within the AWS organization, I worked in the Deep Engines team as a part of Amazon AI. Specifically, I worked on distributed deep-learning model training. The Deep Engines organization was responsible for optimising Apache MXNet, an open-source deep learning framework (competing against the likes of Google’s Tensorflow and Facebook’s PyTorch).

Before joining AWS, I was a CS graduate student at Georgia Institute of Technology. Over the period of 3 semesters at GaTech (Fall 17, Spring 18, Spring 19) I was a Graduate Research Assistant (GRA) at Center for 21st Century Universities involved in research around Education Technology. In my final semester (3rd), I got to work on a Blockchain POC (Proof of Concept) project (Blockchain for Academic Credentials) with C21U. I got to contribute to the Blockcerts repository (MIT’s blockchain for academic project).

Intern in US

Thanks to Georgia Tech’s Co-Op program, I spent my Fall 2018 semester in Palo Alto, California working with Amazon Web Services (AWS) as an SDE Intern. I was lucky to be assigned Amazon AI org and specifically in the Deep Engine team. Over the 4 months (Fall 2018), I could get my hands dirty with mathematics, C++ and Machine Learning (you may call it Deep Learning to sound cool). To be precise, instead of just “using the framework” (import mxnet as mx), I could look at the framework from architecture standpoint; learning it’s front-end and back-end. The name of the framework is Apache MXNet, an incubator project licensed under Apache. It is an open-source Deep Learning framework (akin to Tensorflow, PyTorch, Keras) which boasts high performance and great results along with user-friendly APIs with a multi-language support (to name a few - Python, R, Clojure, even Java APIs). All in all, it was an incredible learning experience to work in the Big 4 Tech company and yet be involved in an open-source project (200+ Github contributions over the year! Yay!)

Over the summer of 2018, I was lucky to be a part of Verizon Connect (previously Verizon Telematics). I got to work on 2 Proof-of-Concepts (POCs) related to automobile/fleet management. First one being “Device to Blockchain” - an effort to prevent odometer fraud and other tampering of vehicular data by removing intermediaries in data measurement and record-keeping. I got to work on Ethereum and Solidity smart contracts, as a part of this project and we were able to demonstrate a product that could potentially help insurance companies, customers, manufacturers and automobile maintenance service agents. Second project was “Predictive Maintenance” (differentiate from Preventive and Corrective). We build a variant of LSTM (Long Short Term Memory) model based on the time-series data about Diagnostic Test Codes (DTC). The model ingested million+ data points to come up with a prediction of whether the car would suffer from any of the top 10 DTCs in the next 30 days - pretty cool feature! Classic multi-class classification problem.

Back to the academics (Grad school)

Complimenting my research work in Edutech, I dived into the realm of Health Care tech and User Experience in the first semester (Fall 2017). Thanks to Lauren Wilcox, I was able to work on a product called Rapport - an interactive website for presenting radiology reports to layman. It involved Natural Language Processing, Model Visualization and Website Development. At the end of the semester, I was able to successfully stitch together BioDigital (human model) with the Radiology reports on the demo website. Rapport CS 8903 - Video

Moreover, I was fortunate enough to be a Teaching Assistant along with a GRA in the first semester of my schooling in USA. Being a TA for Physics course was super challenging. I was responsible for conducting recitations for 3 batches over the semester and taught around 70+ students overall. Revising foundational concepts of electro-magnetism, electric circuits and solving problems on Kirchoff’s and Gauss law took me back to my undergrad whilst exposing me to US teaching system. All in all, it was an insightful experience.

First Foray into the computer science world (Under-grad)

As a part of my 4-year undergrad (Bachelors in Computer Engineering at Sardar Patel Institute of Technology, I was actively involved in research work and applied Machine Learning. Dabbling my hands in the field of Image Recognition and Machine Learning, I researched on Skin Image Recognition using RGB, HSV and YCbCr color models. We successfully detected and differentiated skin pixels from non-skin using a weighted combination of different color models. Our work revolved around training and building the mathematical formula to help improve the accuracy (>90%). Further, I have assisted Nexchanges in building an end-to-end Data Analytics pipeline. Right from Data Extraction using Python libraries Scrapy and Beautiful Soup to Data Warehousing to the final Data Visualization using D3.js, I have learnt immensely thanks to Nexchanges.

O hail Open Source!

Being a firm believer in open science framework and the principle of reproducible work, I actively publish my code on GitHub. Over the past year, I have managed to become a Contributer to the open source Deep Learning framework, Apache MXNet. Being an incubator project under Apache organisation, it allows me to be a part of a vibrant and intelligent open source community. I aspire to be elected as a Committer (next logical step) of the Github repository : apache/incubator-mxnet

For more details, have a look at

Moreover, to catch up with my tweets (I tweet daily about anything interesting I find), Follow @chaibapat

News

Dec 2020	Launched Sagemaker Distributed Data Parallel library for faster, cheaper, efficient distributed training on AWS.
March 2019	Won MIT Sloan's Sports Analytics Conference (SSAC) Hackathon 2019 in the Gamification Division. Represented Georgia Tech. LinkedIn report - WTA Fan Coin: Redefining Fan Experience in Women's Tennis
March 2018	Won Statefarm Challenge at HackGSU 2018. Built an Augmented Reality-based (AR) Insurance App for Statefarm.
August 2017	I began my coursework for Masters in Computer Science at Georgia Tech.
June 2017	Our IoT paper accepted to SSCC, Springer and the camera-ready versions are available on arXiv.

Education

Aug 2017 - May 2019	M.S. in Computer Science (3.625/4.000) Georgia Institute of Technology
Aug 2013 - May 2017	B.E. in Computer Engineering (8.4/10.0) Sardar Patel Institute of Technology, University of Mumbai
Aug 2011 - May 2013	High School
May 2011	Parle Tilak Vidyalaya English Medium School (Mumbai, Maharashtra, India)

Research Experience

Jan 2019 - May 2019	Georgia Institute of Technology, Center for 21st Century Universities Matt Lisle Education Technology - Blockchain BlockCerts
August 2017 - May 2018	Georgia Institute of Technology, HX Labs Lauren Wilcox Health Care Technology and User Experience Rapport - Interactive Patient centered Radiology reports [video]
August 2017 - May 2018	Georgia Institute of Technology, Center for 21st Century Universities Dr. Rob Kadel Education Technology - Data Analysis Performance Analysis of MOOCs vs On-Campus Teaching
June 2016 - May 2017	Sardar Patel Institute of Technology, Dr. Anant V. Nimkar Internet of Things and Security
May 2015 - July 2016	Sardar Patel Institute of Technology, Dr. D. R. Kalbande Image Recognition and Machine learning

Publications

Blockchain for Academic Credentialsy Chaitanya Bapat arXiv 2020 [1] [abs] [pdf] Academic credentials are documents that attest to successful completion of any test, exam or act as a validation of an individual’s skill. Currently, the domain of academic credential management suffers from large time consumption, high cost, dependence on third-party and a lack of transparency. A blockchain based solution tries to resolve these pain-points by allowing any recruiter or company to verify the user credentials without dependence on any centralized third party. Our decentralized application is based off of BlockCerts, an MIT project that acts as an open standard for blockchain credentials. The project talks about the implementation details of the decentralized application built for BlockCerts Wallet. It is an attempt to leverage the power of the blockchain technology as a global notary for the verification of digital records.

Smart-Lock Security Re-engineered using Cryptography and Steganography Chaitanya Bapat, Ganesh Baleri, Shivani Inamdar and Anant V Nimkar arXiv 2019 [1] [abs] [pdf] After the rise of E-commerce, social media and messenger bots, rapid developments have been made in the field of connecting things, gadgets, and devices, i.e, the Internet of Things (IoT). In the fast-paced lifestyle, it is very difficult to maintain multiple keys for tra- ditional mechanical locks. Electromagnetic smart locks are a possible solution to this problem. To connect a smart lock with a key, Bluetooth Low Energy (BLE) protocol can be used. BLE protocol is vulnerable to Man-in-the-Middle (MITM) attack. Ensuring security over BLE is an ongoing challenge. This paper aims to analyze the MITM vulnerability of BLE and develop a possible solution for designing smart-locks with an increased level of security. The observation shows that the combina- tion of Image Steganography and Cryptography helps to overcome the vulnerabilities of BLE protocol.

Human Skin Detection Using RGB, HSV and YCbCr Color Models S. Kolkur, D. Kalbande, P. Shimpi, C. Bapat, and J. Jatakia arXiv 2017 [2] [abs] [pdf] Human Skin detection deals with the recognition of skin-colored pixels and regions in a given image. Skin color is often used in human skin detection because it is invariant to orientation and size and is fast to pro- cess. A new human skin detection algorithm is proposed in this paper. The three main parameters for recogniz- ing a skin pixel are RGB (Red, Green, Blue), HSV (Hue, Saturation, Value) and YCbCr (Luminance, Chromi- nance) color models. The objective of proposed algorithm is to improve the recognition of skin pixels in given images. The algorithm not only considers individual ranges of the three color parameters but also takes into ac- count combinational ranges which provide greater accuracy in recognizing the skin area in a given image.

Teaching Experience

F2017	Georgia Institute of Technology Physics 2211/2212 , TA
F2016	Sardar Patel Institute of Technology Object Oriented Programming Methodology , TA
S2016	Sardar Patel Institute of Technology Structured Programming Approach , TA

Industry Experience

Jul 2019 - Present	Amazon AI, DeepEngine Team, Software Development Engineer
Sep 2018 - Dec 2018	Amazon AI, DeepEngine Team, Software Development Engineer Intern
May 2018 - Dec 2018	Verizon Connect, Software Development Engineer Intern
May 2016 - Dec 2016	Nexchanges Technology Private Limited, Data Scientist Intern
Apr 2016 - May 2016	MatriCS, Research Intern
Jun 2015 - Dec 2015	Quickwork Technologies Privated Limited, Software Developer Intern

Skills

Languages	Python, Hacklang, PHP, Java, C, Bash, C++, CSS, HTML, JavaScript, LaTeX, Make, R, Scala
Frameworks	Apache MXNet, NumPy, Pandas, SciPy, scikit-learn, TensorFlow, PyTorch, Spark, Hadoop, Pachyderm, Android SDK/NDK, Node.js, Kafka
Systems	Linux, OSX

Recent Blog Posts

2018: Year in Review	February 1, 2019
Springer Mountain & Amicalola Falls with OR@GT	June 24, 2018
If I were to build a company right now	June 11, 2018
Interactive Resume using Tableau	March 14, 2018
Makeover Monday Week 3 15 Jan 2018	January 15, 2018

View all

---

Last updated on 2020-01-12