United States

AI/ML Engineering

Master’s Degree

Bachelor’s Degree

Ravi Shankar Kannan

Senior AI/ML Engineer

Machine learning

Unit Testing

Pandas

Python

PyTorch

Docker

Jupyter Notebook

Jenkins

TensorFlow

AI/ML

- Semi supervised learning using Noisy student learning to improve model performance using pseudo labels. 
- Prompt engineering on CLIP model using zero shot learning for new fashion domain data.
- Building ML pipeline for model training with docker. Trained several hundred model with multiple architectures and hyperparameters
- WANDB data and model management integration with LilyAI codebase.
- WANDB sweeps for hyperparameter tuning and optimization.

Building scalable AI technology to save lives and reduce human suffering. 
 - Proven experience in research and implementation of ML models and deep neural networks. 
 - Part of the team that developed end-to-end applications such as Child Abduction Retrieval Technology (CART) on live traffic camera data, Banjo Live Time (BLT) for live real-time signals, vehicle collision detector for intersections (SI).
 - Implemented Stopped Vehicle Detector using real time object detector and Kalman tracker for CART.
 - Research in several areas that include SOTA object detection & image classification approaches for replacing existing models for BLT and CART using deep learning and traditional CV techniques.
 - Research into non-deep learning techniques and traditional computer vision algorithms to speed up CART pipeline. 
 - Deployed CART and several other services using docker on kubernetes cluster and worked on-call for the application support. 
 - Repurposed real-time object detectors like YOLOv3, Fast RCNN, efficientDet and CenterNet for object detection research; Transfer learning pretrained models for building a (vehicle) color detector for CART pipeline. The pipeline also includes several preprocessing and post-processing techniques for video detection system. 
 - Engineered an end-to-end curation system using a third party tool called Prodigy that helped us build an in-house curation pipeline. This helped in data annotation by Banjo's experts instead of outsourcing data and enabled model retraining. 
 - Used time series analysis to analyze opioid overdose reports and to detect anomalies in roadway accidents. Used ARIMA, SARIMA and LSTM.
 - Used NLP techniques to remove/avoid ingesting unsafe text signals into database.

Researched and developed an application to measure the fingernail growth with images using RANSAC, graph-cut, segmentation and deep learning which helped Usana Health Sciences further their research.
Interfaced stereo camera to detect user’s face angle and depth with a treadmill while estimating gait.

Teaching Assistant - ME 1000 - Design of Mechanical Design for Engineering Systems Lab.
Responsibilities included:
Teaching the CAD modelling software - Solidworks, to 100+ students.
CAD modelling and design of fixed and movable systems.
Consultant for the students' final project - Designing a Mechanically powered water pump.
3D printing of various complex designs of water pump and supporting components used in the final project.

Teaching Assistant - ME 1010 -  Computer-Based Problem Solving for Engineering Systems
Teaching students to program on Arduino, C and Matlab

Developed a banking application for World Bank Group aimed at managing banking services. 
We built the service based on a test driven approach using ASP.NET and Oracle for database management.

Ziwei Hong

Manager, Advanced Analytics

yihan zhou

Senior product designer

Ravi Shankar Kannan🇺🇸

About Ravi Shankar Kannan

Expertise

What mentees say