Projects

Microsoft Corp. May ' 20 - Aug ' 20

Continuously monitor the telemetry and trigger automated performance tune up routines, on the data nodes who are maxing out on capacity.
Tech - C#, C++, KQL (distributed systems)

Stony Brook University. Jan' 20 - May ' 20

Extracted structured data from unstructured pdf & html text documents using NLP.
Enhanced the average extraction performance accuracy by 30% overall.
Tech - Python (sci-kit learn, pandas, numpy, sci-py)

Stony Brook University. Jan' 20 - May ' 20

Trained several ensembles of classifiers to perform semantic image classification.
Implemented a successful ensemble of classifier, that is robust against PGD attacks. (GTSRB dataset)
Tech - Python (tensorflow)

Stony Brook University. Sept' 19 - Oct' 19

Implemented SHARK2 algorithm to decode the gestures into typed words.
Optimized the implementation of the algorithm for faster run-time by performing parallel computing operations using Sci-py and numpy libraries.
Tech - Python (sci-kit learn, pandas, numpy, sci-py)

Stony Brook University. Oct ' 19 - Nov' 19

Performed hypothesis testing to determine which user interface is the best among the given options.
Performed t-tests, pairwise t-tests, ANOVA and MANOVA to determine the correct hypothesis.
Tech - R (stats, ggplot2)

Symantec Corp. Jan '17 - Jul '17

Performed dimensionality reduction to reduce the number of attributes from 205 to 139.
Trained a Gradient Boosted Model on the static attributes of portable executable files to detect ransomware at an accuracy of 89.80% with 0.003 FP rate.
The classifier detected the samples of the infamous ‘WannaCry’ ransomware.
Tech - Python (sci-kit learn, pandas, numpy), statistical analysis, XG-Boost, Random Forests and ensembling

Symantec Corp. Jan '18 - April '18

Trained a convolutional neural network, to detect PUAs (potentially unwanted applications), based on screen shots. The model detected the PUAs at 83% of accuracy and 4% of FP rate
Tech - Python (sci-kit learn, pandas, numpy, nltk, keras, tensorflow, seaborn), Windows API

Symantec Corp. Jan '19 - April '19

Designed and developed a distributed software system that collectively convicted similar files based on rules.
Also designed a small protocol for the EC2 instances to communicate among themselves.
Tech - Python (boto3, urllib2, seaborn), AWS (EC2, S3, SQS), MS-SQL

Symantec Corp. April '18 - June '18

Performed data cleaning, dimensionality reduction and feature ranking of the attributes.
Trained an award-winning machine learning model to detect threats on Mac OS proactively. (Symantec L2 award)
Tech - Python (sci-kit learn, pandas, numpy), statistical analysis, XG-Boost, Random Forests and ensembling

Participated in Club Mahindra Data Science Hackathon to predict the expenditure of the tourists.
Secured 242'nd place among 4105 Participants.
Performed data cleaning, EDA, statistical data analysis and built several machine learning models to predict the sales.
Tech - Python (sci-kit learn, pandas, numpy, fastai), statistical analysis, XG-Boost, Random Forests and ensembling, Regression Techniques and PCA, LightGBM
Link (to leader-board handle niranjansa)
Code

Participated in the Innoplexus Sentiment Analysis challenge hosted on AnalyticsVidya.
Performed data cleaning, feature selection, statistical analysis and model building.
Tried the following approaches -
- Bag of words
- TF-IDF
- Use of pre-trained models.
- Word Embeddings
Tech - Python (keras, tensorflow, word2vec, Spacy, nltk, sci-kit learn, pandas, numpy, seaborn, fastai), XG-Boost, LightGBM
Link to final code for the best performing model.

Symantec Corp. Jul '17 - Sept '17

Enhanced a random forest classifier for a 2% gain in accuracy, reduced the FP rate by 10% to 0.006 and reduced the memory requirements by 40%.
Tech - Python (sci-kit learn, pandas, numpy), statistical analysis, XG-Boost, Random Forests, LightGBM, dimentionality reduction, regularization techniques.

Participated in Microsoft's Malware detection challenge. It was my first attempt to solve Kaggle challenge.
Performed a detailed data analysis, feature engineering and model building.
Final model stood at 1044th position out of 2426 participating teams.
Tech - Python (sci-kit learn, pandas, numpy, fastai), statistical analysis, XG-Boost, Random Forests, LightGBM, dimentionality reduction, regularization techniques.
Link to the code of final model.

Walchand college of Engineering May '16

Designed and Developed a music player for Android phones.
Designed and implemented a simple song recommendation algorithm based on fuzzy mathematics to recommend songs.
App had the functionality to share your musical interests with your friends.
Tech - Java: Android Programming (Content providers, services, threads, user interface design, processes, SQLite), Firebase

Symantec Corp.

Page updated

Google Sites

Report abuse