Devesh Datwani

About me

Robotics engineer who's about AI products.
When the internet became a consumer commodity, many shared concerns about their jobs becoming redundant as a result. The same has been said for all revolutionary technologies and tools. And every time, they have generated more jobs than they ended. Anecdotes like these have been said a zillion times already. It's a bit cliche at this point.

I stand on the optimistic side.

My ML/software journey began with my passion for aviation, which led me to build RC planes in high school as a hobby and then pursue studies to be able to build autonomous aerial vehicles. In this pursuit, I studied computer vision, deep learning, autonomous navigation, fluid dynamics, robot perception, robot localisation, and then... NVIDIA and transformers happened, driving my curiosity further in deep learning-based perception systems.
With all my learnings, which are forever on-going, of course, I am enjoying my full-stack AI product journey where I build, test, sell, and ship AI products. It is thrilling to explore what AI and readily available cloud GPU compute is empowering us/me to build. ClosetX and Timelight are the latest personal endeavours outside of my work life, wherein I try to leverage AI to build tools that enhance consumer experience in two different industries.
If you build autonomous vehicles or software that leverages AI-GenAI to enhance the user experience in fashion, real estate decor, insurance, strategy, or any other exciting space, feel free to reach out to me through my email for a coffee chat.

Endeavour Summary

✈️ Flying High with RC Planes
Designed and built RC planes powered by gas and electric motors, experimenting with composite materials and electronics.

🎓 Formalizing My Passion for Aviation
Pursued mechanical engineering and robotics to deepen my understanding of aviation and robotics.

🚒 Robots to the Rescue
Partnered with the Worcester Fire Department to create a unique fire safety evaluator with 2D LiDAR maps from mobile robots to assess indoor fire risks, bringing robotics to public safety.

🏆 Making Waves in Sports Tech
Working with major sports organizations to track objects of interest for data generation and analysis. Operating cloud-based systems to ensure smooth software lifecycles and efficient data storage.

🛍️ AI Meets Fashion
Built ClosetX, a distributed system using vision transformers to recommend compatible fashion items. Brought together after only a year of experience in engineering, AI, and system design to create an innovative shopping solution.


Education

Worcester Polytechnic Institute
Masters of Science in Robotics Engineering
Worcester, MA, USA
2021 - 2023
Computer vision, Machine learning, Robot perception, Project management
University of Mumbai
Bachelor of Engineering in Mechanical Engineering
Mumbai, MH, India
2014 - 2018
Machine design, Structured programming approach C/C++, Theory of machines, Operations research, Heat transfer, Fluid dynamics


Highlighted Experiences

Bolt6
Development & Operations Engineer
Charlotte, NC, USA
2023 - present
Product development & operations, Kubernetes, AWS, Deep Learning
Human inspired robotics lab
Research assistant
Worcester, MA, USA
2023
Augmented reality, Task affordance, Unity, Shared control
Nurturelabs
Software developer intern
Remote
2021
Django, FastAPI, Postgress, Python, Business lead generator
Admatazz
Data Analyst
Mumbai, MH, India
2019 - 2020
Product development, Re-marketing tools, Web scraping, Object detection,
Air India engineering services
Intern
Mumbai, MH, India
2017
CFM-56, GE90, Engine overhaul, Operations, Non invasive testing


Projects

ClosetX
Microservices, Full stack deep learning, AWS
Fire safety evaluator
Lidar maps, PCA, Image processing, ROS, State machine
Deep segmentation with attention
Unet, Pytorch, CUDA
Fine-tuning YOLO for traffic sign detection
Tensorflow, Data augmentation, Object detetion, Pre-training
Trends on the map
Flask, MYSQL, Twitter API, AWS, Restful API
Structure from motion
PnP, RANSAC, Epipolar geometry, 3D triangulation
Occupancy grids with bayesian algorithm
Log odds, Voxel filtering, SLAM
Image stitching with deep homography network & closed form RANSAC
RANSAC, Harris corner detection, deep learning
Airflow induction with plasma actuators
Ion thruster, Corona discharge, Flow control, Di-electric barrier discharge
Deep pose estimation with cascaded convolution neural networks
Cascaded VGG16, Regression model, Keypoint estimation