Suyash Srivastava

Email: suyash.cn.srivastava{at}gmail{dot}com LinkedIn: linkedin.com/in/suyashcsrivastava Github: github.com/catch-n-release

I am learning how a machine learns.

There is no denying that technology has seen massive shifts in recent times, making way for a host of possibilities that were unimaginable just two decades ago. I see myself contributing to this rapid transformation of technology. I desire to be a technology leader like Sir Timothy Berners-Lee when he created the World Wide Web. I want to emulate Steve Wozniak's drive for innovation when he built the personal computer and Linus Torvalds' intent when he developed the Linux kernel. Like them, I wish to dive deeper into the problems and processes that give birth to new technology.

As a part of the Machine Learning capability of Deloitte Consulting USI Studios, I have almost three years of experience in designing, implementing, and delivering computer vision and natural language processing solutions using machine learning and deep learning. I applied my skills in developing image processing solutions for Deloitte's in-house document processing tool and also for Waste Management's AI arm for processing totter videos. Additionally, I am efficient in exposing these core components as services by developing scalable and robust microservices. And subsequently deploying such service on multiple cloud platforms such as AWS, Azure, and Heroku to name a few. As a consultant, I have also chaired teams of developers along with training and bringing them up to speed in projects.

When I am not training models or developing, you'll find me drumming to alternative rock, progressive metal, and indie songs. (Scroll down for videos!) I also like to hike and go camping in the woods.

By combining the finest of graduate learning from Carnegie Mellon University with my prior professional experience, I aim to go back to the industry to work in the field of AI and ML. The copious amount of growth possibilities is everything I seek in order to take my dream flight beyond the clouds while being noticeably conducive to the tide.

Fast ramp up on new areas and technologies.

Ability to coordinate across multiple teams and verticals.

Analytical and debugging skills.

Open source contributor.

Profound knowledge in Convolutional Neural Network (CNN), Computer Vision, Natural Language Processing (NLP) and Application Programming Interfaces.

Experiences

Palo Alto Networks

Software Engineer Intern

May 2022 - Aug 2022
Deloitte

Consultant

Jun 2021 - Aug 2021
Deloitte

Business Technology Analyst

Oct 2018 - Apr 2021

Education

Carnegie Mellon University

Master's, Information Systems Management

Aug 2021 - Dec 2022
University of Pune (Savitribai Phule Pune University)

Bachelor of Engineering, Electronics and Telecommunication

Aug 2014 - May 2018

Awards & Recognition

Deloitte

Applause Award

Nov 2020
Deloitte

Spot Award

Jan 2020
AICTE

AICTE Grant

Apr 2018 Received a grant-in-aid of INR 3,00,000. For post-development activities and further research on the project.
KPIT Sparkle

Finalists (Ranked 7th)

Feb 2018 Used Image Processing, Convolutional Neural Networks and Remote Communication to develop an advanced overtake assistance system. It helps execute a safer overtake manoeuver by relaying live dashboard camera video from preceding car to following car. Project Report Project Demo
Smart India Hackathon

Winner

Apr 2017 Built a Smart Image Processing solution to identify violators over-speeding at traffic signals using Object Detection and CNN for Automatic Number Plate Recognition on video feed provided by traffic cameras Covered on NHIDCL

Skills

Languages

Python, C, C++, Java

Technology Trends

OpenCV, PyTorch, Keras, Spacy, TensorFlow, SQLAlchemy, FastAPI, GCP, Redis, AWS:LEX, KENDRA, API GATEWAY, LAMBDA, SQS, SNS

Workflow

Bitbucket, Git, Agile and PMM

Database

Neo4j, PostgreSQL, MySQL, DynamoDB

DevOps

AWS, Heroku

NOTEBOOKS

Tiny GoogLeNet on CIFAR-10

Non-sequential tiny version of GoogLeNet network shows 90% classification accuracy on CIFAR10.

Keras, scikit-learn, OpenCV, Numpy, Matplotlib, Python - Notebook

Network Ensemble

Ensemble of TinyVGG shows increased classification accuracy on CIFAR10.

Keras, scikit-learn, OpenCV, Numpy, Matplotlib, Python - Notebook

Regularization Techniques

Regularization technique improves TinyVGG accuracy from 61% to 71% on Flowers17.

Keras, scikit-learn, OpenCV, Numpy, Matplotlib, Python - Notebook

TinyVGG on Flowers17

Training TinyVGG network on Flowers17 Dataset and obtaining 61% accuracy.

Keras, scikit-learn, OpenCV, Numpy, Matplotlib, Python - Notebook

Transfer Learning: Networks As Feature Extractors

Using state-of-the-art networks like pre-trained VGG16 as a feature extractor for obtaining 92% accuracy on CALTECH-101.

Keras, scikit-learn, OpenCV, Numpy, Matplotlib, Python - Notebook

Portfolio Analysis

Evaluating portfolios and their benchmarks, with documentation and manual calculations to illustrate all underlying statistics.

Yahoo Finance API, Python - Notebook

Waste Management

Solution for determining the overage and contamination of dumpsters using ML and Image Processing.

Cloud Formation, AWS ( LAMBDA, SQS, DynamoDB, SNS, S3), OpenCV

Contact Matching

Singular Asynchronous Contact Matching & Updating service for bulk of ranging up to 100k records

Heroku, Flask, PostgreSQL, Celery, Redis, Salesforce

SmartTalk

A closed Domain Question Answering tool using BERT (Bidirectional Encoder Representations from Transformers)-A state of art language model for NLP

CDQA, BERT, Neo4j, Docker, FastAPI, GCP

DocuEdge (SmartBox)

A Machine learning tool to classify any document to specific types, by a trained model. Further extracting relevant information from classified documents using Natural Language Processing

OpenCV, Neo4j, Docker, PostgreSQL, Redis, Faiss, Milvus, FastAPI, AWS




CMU IMO