About Me
Hey! I am currently an engineer at Amazon, India working in the Transporation Technology Org. I completed my undergraduate degree in 2021 from PES University, Bangalore, India with a Bachelors in Technology in Computer Science and Engineering and a specialisation in Data Science.
My work at Amazon operates on a massive scale. We're responsible for generating the tracking IDs, labels, manifests and invoices for 35% of Amazon's worldwide traffic.
My areas of research interest are mainly NLP - speech & text, applied data science and ML, information visualisation and statistical methods for data
Work Experience
Amazon, India
Programmer Analyst (Software Development Engineer)
January 2021
Working in the Transportation Technology Org, I have worked on migrating multiple software services from legacy platforms onto newly created architectures. I took care of shadow mode implementations as well, making sure there is no impact in Production. I am responsible for onboarding multiple third party carriers onto the Amazon network by designing, planning and finally implementing their specific requirements to integrate with our systems. We ensured a smooth 'Prime Day' across 2 years with no downtime.
Happiest Minds Technologies
Software Development Engineer - Intern
June 2020 - August 2020
Built a drone based surveillance system to detect objects, track and localise them in 3D. The solution involved comparing the working of two models - Retinanet and a RCNN on Meta’s Detectron dataset, to find the most efficient model for the company's use case. Post that, we built a system to track the objects, and finally pinpoint them.
PACE Global Consulting Services
Software Development Engineer - Intern
June 2019 - August 2019
Worked on automating a part of the recruitment process for the company. Used a random forest classifier to set up an initial screening process for the resume. The model takes into account the candidate's prior experience, objective and academic performance and then suggests whether the job opening under consideration will be suitable for the candidate or not. Also built a web scraping algorithm to scrape resumes off job application websites.
Sasken Technologies
Software Development Engineer - Intern
June 2018 - August 2018
Contributed to the development of KenSEWA, an application developed by Sasken Technologies which works as a database management system and finance automation tool for women in rural areas. Worked on developing the UI for this platform using HTML, CSS and JavaScript for a seamless user experience.
Skills
Frontend
Backend
Programming
DevOps
Select Research Projects
Rumour Detection and Veracity Verification
Early detection of rumours and assessing their veracity is a key problem. The learning mechanism analyses the socio-linguistic data and the features of the social graph of the Tweet. We first perform a graph and text based rumour detection, followed by classifying the stance of the Tweet. We then finally predict whether the Tweet in question is True or not.
Cloud Computing - DbaaS - Container Orchestrator
This simulates a ride-sharing application. The orchestrator exposes the read and write API endpoints. Out of all the worker containers, one of them is elected as the master. We have used queues for message exchanges. We scale up/down the number of containers based on the number of incoming HTTP requests. We use Zookeeper for fault tolerance.
Product Expiry Notification System
Built a product expiry notification system to scan and predict the expiry date of certain household essentials. Used OCR, Android Studio, an ANN and an ANOVA test to implement the solution. This used an adam optimizer, an MSE as its loss function and was finally evaluated on its cosine proximity. We then used a Graph Neural Network to compare functioning of models.
Lie Detection System - Eagle Strategy
This has 3 parts, camera for facial features, recording device for voice and IOT devices for body movements. The truth value of every line in a conversation is captured. Based on this, we trained our deep neural network to classify the given data as a truth or lie. I used the Eagle Strategy in place of the generic back propagation algorithm used in a neural network.
Naive Bayes Classifier
This classifies whether a person is a democrat or republican based on 16 binary attributes and 2 classes. I estimated the accuracy of Naive Bayes algorithm using 5-fold cross validation on the dataset given. Implemented the model from scratch, without using any libraries. A decision tree, using the ID3 algorithm, has also been used to classify data and compare the performances of the two classifiers.
C++ Compiler using the C Language
This is to build a compiler for the C++ programing language, using the C programming language. The compiler handles 'do', 'do-while' loops and 'if' conditional statements. The phases of this project include the lexical, syntactic, semantic analysis. Post that, we work on the intermediate code generation, optimisation and finally generate the target assembly code.
Stock Market Analysis and Prediction
This helps users make decisions when investing in the stock market. On the basis of data collected over a long period of time, this uses support vector regression to predict the upcoming month's price of a stock. It uses Flask and jQuery for the backend, and Angluar for the front-end requirements. An appleaing and easy-to-use UI has helped improve the overall UX.
Emotion Detection and Classification in Speech
Used the RAVDESS dataset and a CNN to build a system that detects the emotion in audio. Classifies the speech into 3 emotions - calm, happy, sad & happy. Used the librosa lib in Python, and extracted the MFCC(Mel Frequency Cepstral Coefficient) from the samples. I used the softmax and rmsprop activation functions and a batch size of 32.