Resume
Sai Adithya Reddy Chinthalapani
Student from Arizona State University with a Master’s in computer science (Big Data specialization).
Education
Arizona State University 2019 – 2021
Vellore Institute of Technology 2015 – 2019
Skills
- Programming languages: Java, Python, Ruby, JavaScript, SQL, AWS, OpenCv, Scikit-Image, Scikit-learn, Keras, Pandas, Numpy, Mongoose, Ruby on Rails, React, Express, Django, MySQL, PostgreSQL, Git, unix terminal
Certifications and Apps
- AWS Certified Solutions Architect Associate SAA-C02
- Developed an android game using GDevelop and Cordova plugin which got 15000 installs.
Hackathons
Action.ML Hackathon - Fox
- Trained the model on AWS Sagemaker by getting data from s3 bucket hosted by FOX.
- Experimented by finetuning models like RESNET and VGG pretrained models in Keras to classify sports scenes
Android Hackathon – Dev Savant
- Developed a quiz app for cricket enthusiasts that randomizes questions based on selected difficulty level in 24 hours.
Experience
Global Launch, ASU : Student Video editor
- Produced videos by following instructions form professors and delivered them within deadlines.
- Developed ffmpeg scripts that automated parts of video editing and saved 100’s of hours in production time.
Data Scientist intern - Verticross India Pvt.Limited
- Created visualizations to find anomalies and missing data using matplotlib and plotly.
- Performed time series analysis and forecasting on data collected from substation meter readers using keras.
- Transformed data using SQL queries, psycopg2
- Wrote several scripts to for data simulation, to calculate performance metrics and to analyze data using numpy, scipy and pandas.
Android Developer Intern - Grepthor Software Solutions pvt. ltd
- Developed an application that connects volunteers and managers using an android application.
- Developed splash screen, authentication features and displayed data, images using Picasso library.
- Followed the material design principles for design and connected to the APIs for data using volley.
Projects
- Implemented different versions of k-means algorithm and Naive Bayes algorithm from scratch using numpy and compared accuracy of both algorithms.
Scalable image classifier – Aws, Python/h3>
- Deployed a deep learning classifier and front-end app on AWS using flask to classify the input image.
- Scaled the app by implementing custom autoscaling application using SQS, S3, EC2 in AWS.
- Detected anomalous activity in videos with an accuracy of 85 % for 2 classes and 55% for 5 classes.
- Used FFmpeg to extract images from video data and Keras for fine-tuning deep learning models.
- Collect large amount of text data using data scraping tools and manual methods.
- Used novel methods to generate required data using pandas and spacy.
- Finetuned the huggingface transformers BERT model on collab to achieve an accuracy of 69%.
- Built a garbage collection robot using robot car chassis, motors, camera and ultra sonic sensor.
- Scraped garbage photos data from internet using beautiful soup.
- Trained a garbage detection model using Keras and wrote RaspberryPi scripts to handle all the robot components and deployed the deep learning model on it.
Big Data - Geo Spatial Data Hotspot Analysis
- Analyzed NYC Taxi data by implementing range and distance queries in order to identify statistically significant spatial hot spots areas using Apache spark.
- Reduced CPU and memory utilization by up to 50% by distributing the load across Hadoop cluster on AWS.
Machine Learning - Gesture recognition using personalized page rank
- Generated files to store gesture data at different levels of abstraction using statistical features and dimensionality reduction with help of pandas, sklearn, scipy and custom functions.
- Visualized and analyzed the data using heatmaps and charts with libraries like matplotlib.
- Implemented personalized page rank and used it with generated data to recognize gestures.
- Visualized data and uncovered patterns in the sales dataset using Tableau.
- Performed sentiment analysis using nltk on comments for each product to visualize the customer sentiment.
- Retrieved images closest to the given medical image to aid in diagnosis. Project based on peer reviewed publication.
- Used feature engineering, dimensionality reduction and machine learning techniques with opencv, scikit-image and keras to achieve an accuracy of upto 60% and precision of 0.55.
- Implemented different noise detection and reduction techniques using opencv and C++ based on multiple peer reviewed publications to find out the best performing method.
- Calculated different quality estimation metrics to find the technique producing image with least noise using matlab.
- Improved the speed of computation using parallel processing with OpenMp in C++.
- Automated parts of video editing pipeline using FFMPEG at global launch to save many hours in production time.
- Produced result 100x faster when compared to traditional methods.
- Performed primary data visualization and analysis using numpy, matplotlib to understand the data and then selected relevant features based on different methods.
- Classified the data points between haze and fog, predicted temperatures using different machine learning and dimensionality reduction techniques using scikit-learn.
- Implemented fake currency detection method during introduction of new currency in India.
- Researched multiple peer reviewed publications on different currencies to implement an effective method to detect fake currency using image from normal camera using Matlab libraries.
Database - Distributed database implementation
- Implemented round robin and range partition using PostgreSQL and psycopg2.
- Implemented range and point queries.
- Implemented parallel sort and parallel join.
Database - Google like Big Table Implementation
- Implemented low level code for Map insert, Batch insert for Bigtable and improved its efficiency using btrees.
- Implemented indexing, querying, row sort and row join operations on Bigtable database by modifying minibase code in Java.
Geo Spatial Data Hotspot Analysis
- Analyzed NYC Taxi data by implementing range and distance queries in order to identify statistically significant spatial hot spots areas using Apache spark.
- Reduced CPU and memory utilization by up to 50% by distributing the load across Hadoop cluster on AWS.
Machine Learning - Predicting meal/no meal data from CGM data
- Extracted different types of features from the given time-series readings and selected relevant features using forward, backward selection.
- Predicted the meal/ no-meal from the time-series data with an mean accuracy of 70% using scikit-learn, numpy and pandas.