I'm a full-time Learner and a Seeker.
I aspire to build a system to provide all of us; not an equal opportunity,
but the one that enables each one of us to flourish exquisitely.
How is a garden mesmerizing; if all the flowers grown in it, are the same?
Methodical Machine Learning Engineer with a cumulative experience of 4 years.
Skilled problem-solver with unique perspective in solutions, focused on answering why do we do what we do while tackling daily problems.
Jan 2023 - Present
Worcester, MA, USA
Got selected for being a TA for the Computational Data Intelligence course at WPI along with the tuition scholarship.
Aug 2022 - Dec 2022
San Francisco, CA, USA
Trained a self-supervised One-shot learning based Head Segmentation model with 0.75 Dice similarity score.
This model will help deliver MRI-based treatments for brain-related illnesses.
It removes the current dependency on CT scan and prevents the risk of cancer due to ionization.
Worked on a high-impact project changing more than 1 million lives by helping detect stroke in the first response.
Auto-generating labels by one-shot data augmentation for MRI segmentation dataset, and implementing UNETR, a Vision Transfomer (ViT) based model.
Also worked on a time series transformer model for stroke detection.
May 2022 - August 2022
Implemented Graph Machine Learning and GPT, T5, Transformer based algorithms to large Knowledge Graphs for automatic multi-hop Question Generation.
This helps to generate 100s of meaningful questions from every single document which further enhances the document understanding for Question Answering.
Researched on applying Graph Machine Learning algorithms to large Knowledge Graphs
for automatic multi-hop Question Generation.
August 2021 - May 2023 (expected)
Worcester, MA, USA
My area of research is in applying Machine Learning and Deep Learning techniques to various problems in Education, Medicine, Transportation, Autonomous Vehicles, and Robotics.
February 2021 - July 2021
Created a pipeline of OpenCV data augmentation, model-assisted data cleaning and labelling to qualify dataset to train Efficientnet models which improved classification accuracy by 20% of frequently varying image designs into 500 classes.
January 2019 - August 2020
Designed and implemented end-to-end ML systems for Global Returns Forecasting,
Dynamic Capacity Planning, etc.
Increased inference performance by 70% and decrease running cost by 20% for
deployment code by bootstrapping.
Implemented parameterized code to enable Neural Architecture Search (NAS)
driven by JSON config.
Mentored an intern for 4 months while implementing EDA and ML data flow
pipelines for Anomaly Detection.
May 2017
Bengaluru, KA, India
Gave a talk on Essence of Linear Algebra in IISc Summer School 2017 to
80 students from across India.
Inspired by the graphical intuition behind the far-reaching applications of LA concepts.
Here's the link.
Thanks to the great videos on the 3Blue1Brown YouTube channel.
August 2016 - December 2018
Bengaluru, KA, India
I studied various courses in the Intelligent Systems stream of the Computer Science and Automation department. Relevant courses: Machine Learning, Deep Learning, Natural Language Processing, Reinforcement Learning, Game Theory, Probability and Statistics. Contributed ideas to the Intelligent Traffic Lights System's project.
August 2010 - June 2014 Pune, MH, India
I studied Computer Science in the D. Y. Patil College of Engineering, Akurdi. I gained first-hand exposure in "Engineering" software solutions for real world problems from scratch. I was exposed to inspiring ideas while doing projects in different frameworks starting from C++ console programs to VB scripts to Web-based projects.
January 2022 - May 2022
Worcester, MA, USA
Trained RL agents using DDPG algorithm to handle mapless navigation in simple and
obstacle filled environments.
Designed the state space and reward function to help the RL agents learn to take efficient
actions and reach goals without any collisions. The solution is scalable, dynamic,
and has minimal manual intervention.
January 2022 - May 2022
Worcester, MA, USA
Our motivation for this project was to improve the offroad path detection performance, and
make it light enough to deploy the trained model on edge devices and fast inference.
And thus, progress towards autonomy level 5 in difficult terrains.
We trained various Deep Learning architectures like PSPHead, ENCHead,
with ResNet pretrained backbone.
We also finetuned the hyperparameters and also met the state-of-the-art performance
on the RUGD and Yamaha dataset.
August 2021 - Dec 2021
Worcester, MA, USA
Worked on a very important and challenging dataset of brain MRI scans downloaded from a
Kaggle competition.
We tried a one-shot atlas based data augmentation and
ensemble methods to improve DNA methylation classification.
Trained 3D CNN and Resnet50 models on 4 different types of brain MRI scan data.
Created an ensemble of these models to increase the methylation classification
performance by 4%.
August 2018 - Dec 2018
Bengaluru, KA, India
This project is done in collaboration Research and development team at Siemens India.
We worked on demonstrating the effectiveness of the Deep RL solutions to be better
than human intuition based Fixed Signal Time controls.
Setup PTV Vissim and Aimsun for simulating the traffic environments based on
Bengaluru's traffic analysis.
Ideated traffic density representation which helped the A3C based RL agents to learn
to create green corridors.
Achieved considerable results wherein Average Waiting time for vehicles is
reduced by 1 minute saving
This contact form is not fixed yet. Please email me directly at nitai.shreedhar@gmail.com and include "isWebVisited=True" in your email.
Worcester, MA
Oakland, CA
US
nitai.shreedhar@gmail.com
Kindly, email for phone number.