Shonit Gangoly | Software Developer

Hi, I'm Shonit Gangoly.

A

Self-driven, passionate programmer with a curious mind. I love solving complex and challenging real-world problems.

About

Dedicated software developer with a strong focus on building scalable systems and optimizing backend services. With expertise in Python, Javascript, and cloud platforms like AWS and GCP, I excel in designing efficient APIs, automating workflows, and integrating machine learning solutions to enhance system performance and reliability.

Experience

Selectiva Systems

Software Engineer

Developed and deployed a predictive maintenance model using PyTorch.
Designed and implemented RESTful APIs within a service-oriented architecture (SOA) framework.
Expertly configured cloud functions to automatically trigger and ingest files from GCS into BigQuery.
Demonstrated strong proficiency in Python scripting, developing robust and efficient scripts.
Acquired thorough knowledge of Google Cloud authentication mechanisms, ensuring secure and seamless access to cloud resources and services.
Developed Apache Airflow DAGs using Composer, incorporating Python Operators and Bash Operators.

Jan 2024 - Present | San Jose, California

Blaze AI (YC W22)

Software Engineer

Orchestrated the deployment and management of staging and production clusters on AWS ECS.
Implemented advanced machine learning algorithms such as gradient boosting and neural networks.
Designed and implemented multiple CI/CD pipelines using Github Actions and Bash Scripts.
Set up efficient data processing pipelines with Apache Airflow, improving data management and processing speed.
Created pipelines that boosted the user base of the Twitter Audience Builder feature from 2M to 20M.
Implemented NodeJS scripts to allow seamless integration of backend APIs.
Collaboratively developed robust UI using React, Next.js and Material UI, enhancing user experience.
Optimized data preprocessing and feature engineering pipelines using Pandas and NumPy.
Supported and managed incident responses, conducting blameless postmortems to continuously improve system stability.

Jan 2023 - Jan 2024 | San Fransisco, California

UT Arlington

GTA Data Science(DATA3401)

Responsible for on-going development and design of the curriculum that includes Applied Algorithms, Data Structures, Regression, Classification
Completed hands-on review and assessment of coursework submitted by 35 students.
Supervised practical work along with advising students on skills, methods and techniques to assist the transfer of knowledge.

Oct 2021 - May 2022 | Arlington, Texas

IBM

Software Developer

Devised a Python-based NLP assistant for IBM clients to automate customer support with tailored responses.
Engineered custom Chatbots utilizing IBM Watson to enhance internal workflows for calendar scheduling.
Implemented Decision Trees/Random Forests models to increase response efficiency of NLP assistant by 30%.
Managed large-scale data processing with DataProc, utilizing PySpark and Spark Streaming Jobs.
Implemented NodeJS scripts to allow seamless integration of the NLP assistant with backend APIs.
Employed NoSQL databases for flexible data management.
Designed multiple ML CI/CD pipelines through Github actions to accelerate code deployment by 35%.
Enhanced user support systems and conducted post-incident analysis to refine service responses.

May 2018 - July 2020 | Mumbai, India

NDS Global

Android Developer Intern

Responsible for building Java based android application that uses forms to collect input data and store in Firebase DB
Significantly improved the response time of application from 2 sec to 0.5 sec by using Titanium SDK as backbone.
Assisted team members with several bug fixes on other android applications
Tools: Java, Android Studio, Appcelerator, Titanium SDK

June 2017 - July 2017 | Mumbai, India

Projects

                
Vision
                  Performing Image Segmentation on pupil and iris.
                
AccomplishmentsDesigned and developed custom U-Net architecture that performs semantic segmentation on masked
                    images of pupils and irises to find their distances.
Created programs to mask the images to feed to ML model

Semantic Segmentor
                  Performing Semantic Segmentation on Objects.
                
AccomplishmentsDesigned and developed custom U-Net architecture that performs semantic segmentation on street
                    objects.
Achieved 40% better performance over stock model and similar models such as FCN32
                    and DeepLabV3

U-Net on Brain MRI images
                  Using U-Net to detect tumors in Brain MRI images.
                
AccomplishmentsDeveloped python application that performs Deep Learning Image
                    Segmentation on Brain MRI images
Implemented custom Thresholding, Shearing, Rotation to highlight the Region of Interest in
                    the image. Achieved 82 % accuracy calculated by Dice, Jaccard Coefficient parameters

Tweet Sound of Money
                  NLP Analysis of Tweets and correlation with stock market.
                
AccomplishmentsDeveloped a custom regression technique for correlation
                    between the tweets and SnP 500 data on a time series graph
 Used NLTK Vader Sentiment to perform sentiment analysis on tweets
                    to find positive and negative sentiment ratios.
Built 4 Machine Learning models LSTM, SVM, Naive Bayes, Random Forest and
                    achieved 80 % accuracy in regression analysis.

Object Detector & Classifier
                  An object recognition app using Deep Learning
                
AccomplishmentsTools: TensorFlow, Python, Jupyter, OpenCV
Use of Deep Learning algorithms such as Faster Region Based Convolution Neural Networks, Single
                    Shot Detectors and You Only Look Onc
A comparative study between all these algorithms using TensorFlow
Object Detecting and Recognizing software using these comparisons

Cloud Based File Storage System
                  An online cloud based file sharing and storage system.
                
AccomplishmentsTools:Java, ReactJS, Springboot, AWS
Developed a AWS cloud-based file sharing and storage
                    system that incorporated multiple users concurrently. 
Created custom File-Sync, Delta Sync, Suspend/Resume and Sync Status
                    functionalities. Utilised thread management to supplement the functionalities. 
Implemented Client/Server side modules and Block
                    based storage mechanisms for file transfer and storage.

The Push
                  An Android based application geared towards Women's Safety.
                
AccomplishmentsTools: Java, XML, Android Studio, Google Firebase
Emergency application to counter violence against Women
Using Geo Location Listener and Message Creator it sends an automated help message with the
                    current location of the user to two pre-selected emergency contacts at the press of a button
Contains an Alarm System that can be used to alert people nearby

Immunity Chat Application
                  A chat application that does not store history.
                
AccomplishmentsAn anonymous identity-based chat application
It disregarded the use of back-end servers to store messages hence all information shared was sent
                    and read real time using only JavaScript and NodeJS

Titanic Kaggle Data Mining
                  Predicting survival rate of passengers onboard the Titanic Ship.
                
AccomplishmentsUse of Random Forest Classifier and K Nearest Neighbours Classifier to predict the survival rate
                    of passengers onboard the Titanic ship
Submitted the prediction to Kaggle
Random Forest gives a score of 77.511%
KNN Classifier gives a score of 81.818%
Visit Medium Blog link to learn more about the project

Convolutional Neural Net on CIFAR10
                  Building a model that does image classification on the CIFAR10 data set.
                
AccomplishmentsUse of Convolutional Neural Networks on CIFAR10 data set to classify images
Employs three different models each with different amount of layers and batch sizes
Best model attains an accuracy of 82%
Visit blog to learn more about project

Overfitting on Polynomial Regression
                  Visualization of Over fitting using Polynomial Regression.
                
AccomplishmentsVisualization of Concept of Over fitting
Employs regression to see how different orders of polynomial functions fit on our model
Uses L2 regularization to overcome over fitting
Visit blog to learn more about project

Naive Bayes Classifier
                  Working of Naive Bayes Classifier.
                
AccomplishmentsNaive Bayes Classifier on Yelp Review Dataset
Shows the effect of cross validation and smoothing on Naive Bayes

Data Mining Project
                  Predicting Diabetes in patients using K-Nearest Neighbors and Decision Tree.
                
AccomplishmentsDiabetes is one of the most prevalent disease among people
I wanted to create an application that would help users get an estimate prediction
                    whether they could have diabetes or not
This would greatly help people in making some lifestyle changes to combat the disease
Created prototype UI for application. Screenshot in Proposal page
Visit blog and check Project proposal to learn more