Sai Aswath

wix Automation

Developed a powerful WIX Automation tool by integrating Rasa chatbot, streamlining processes and enhancing user experience through seamless interactions

Talk to Data (Text2SQL)

I led the creation of a top-tier Text2SQL product using in-house Large Language Models (LLMs) like EleutherAI-gpt-j-6b, Falcon-40b-instruct, and LLama2. I established a seamless pipeline for natural language queries to intricate databases, optimized with advanced vector embedding and databases. I excel in fine-tuning LLMs using state-of-the-art techniques like LoRa, Q-LoRa, and SFT Trainer, elevating model performance for accurate outputs. Additionally, I enhanced LLM deployment efficiency through advanced quantization, maximizing GPU resources and real-time responsiveness.

PDF Q&A Bot

smart PDF Q&A bot with Langchain and Llama-index tech. Built a seamless pipeline for PDFs, including OCR and table extraction. Used Elastic and Semantic search combo for precise, contextual answers.

Data compression project (General Motors USA)

• Data compression using Spline regression,EDA on Huge dataset and achieved a 90% compression and with 98%accuracy

Driving Behavior Analysis & Speed Bump Detection

• Collected IMU sensor data & Converted Data into the fixed frame of reference and filtered Noise, Used K-Means to label data, Peak detector for cross-validation and 1-D CNN for Training Model and obtained Accuracy of 98%

UI Test AID Audio detection feature integration (AKKODIS France)

• Used MFCC to extract features from the audio signal and created a model using DTW (tools:-python,Librosa,PyTorch)
• Developed Multiple POC (Proof of Concepts) on image Classification & Computer vision using (YOLOV7,
  ResNet50, and InceptionV3)
• Created the chatbot to automate the HR activities in the company.

• Revolutionized the yoga experience by implementing three layers of AI for asanas classification, positional feedback, and pressure analysis.
• Created applications on embedded systems using CM3 technology, showcasing a mastery of hardware and software integration.
• Pioneered custom computer vision algorithms,enriching Yogifi's offerings by Integrated positional feedback through computer vision for enhanced Yoga Asanas practice,Integrated gym workouts with media pipe technology, developing a sophisticated rep-counter that blends seamlessly with human activity recognition
• Automated app testing using Appium, ensuring the flawless performance of Yogifi applications,Integrated mat testing with the Yogifi app, providing a comprehensive quality assurance process.
• Managed production and staging environments of Yogifi Mat OS, Computer Vision OS, Spearheaded the management of Google Cloud Platform services for Yogifi, maintaining a commitment to excellence.