Shrivarshan K Email & Phone Number

Chennai, TN, IN

Chennai, Tamil Nadu, India

Working on a survey paper - "Small Language Models (SLMs): A Comprehensive Survey"

Bengaluru, Karnataka, India

• Key member of Demand Forecasting Team, leading development of Version 3 forecast engine and process automation
• Engineered V3 forecasting model using NHITS, NBEATS, RNN, and GRU architectures, designing innovative bucketing strategies for accuracy and scalability
• Achieved 16% reduction in Mean Absolute Percentage Error (MAPE) by expanding hyperparameter space and optimizing validation periods per bucket, driving an estimated $0.73M in revenue impact
• Enhanced operational efficiency by 65% through automated workflow integration with Azure, reducing processing time from 35 hours to 12 hours

Chennai, Tamil Nadu, India

A key member of the founding team, instrumental in idealizing the outline of Trulio.io and driving AI-focused initiatives.Trulio.io, a groundbreaking SaaS product revolutionizing search automation on top e-commerce platforms like Amazon, Flipkart, Myntra, and more.Leveraging latest cutting-edge technologies like Gemini 1.5, roBERTa for precise product.

Canada

• Contributed to projects at Transportation Lab, SALMALIS, under Prof. Suliman Gargoum, focusing on traffic safety using advanced deep learning techniques. Worked on a journal paper “Enforcing Traffic Safety: A Deep.
• Processed 100 videos (20,000 frames) of Indian traffic from the AI City Challenge 2023 dataset, covering seven distinct classes
• Conducted model comparison (R-CNN, SSD, YOLO); implemented YOLOv8 for motorcyclist helmet detection, achieving 0.91 F1 score
• Improved F1 score by 2.5% through DCGAN-generated synthetic images and augmentation to address class imbalance
• Applied Test-Time Augmentation (TTA), leading to an additional 5.5% improvement; research findings published

Chennai, Tamil Nadu, India

• Worked under Dr. Swaminathan, in collaboration with Team KAATRU, led by Prof. Raghunathan Rengaswamy.
• Spearheaded the acquisition of high-resolution PM2.5 data for Ambattur city at 250m spatial and 1hr temporal intervals through the deployment of cutting-edge IoT sensors.
• Introduced a novel agent-based modeling approach, optimizing parameters with an Ordinary Least Squares method and achieving a remarkable MSE of 0.616.
• Proficiently utilized GeoPandas for land-use data and the Google Maps API to extract traffic data for each of the 100 agents.
• Elevated the modeling equation by incorporating crucial factors such as land-use and traffic data, leading to a substantial reduction in the model's MSE to an impressive 0.535.

Master'S Degree, Data Science

Bachelor'S Degree, Metallurgical And Materials Engineering

High School Diploma, Computer Science