SANCHAY REWAR

Experience

• 

  Larsen & Toubro

  May 2019 - Jun 2019

  Summer Intern

  I did 45 day internship in Larsen and toubro and learned about the working and construction of a WATER TREATMENT PLANT

• 

  Chegg expert

  Apr 2021 - Aug 2021

  Student
• 

  Telerad Tech

  Jul 2023 - now

  AI Engineer

  -Designed and implemented a deep learning architecture for the accurate detection, localization, and quantification of aortic aneurysms in CT scans, achieving 94.5% sensitivity and 96% specificity. This solution significantly enhances efficiency and reduces turnaround time for radiologists in emergency settings.-Designed, developed, and deployed an architecture capable of running multiple deep learning AI models. This product serves as an orchestrator, efficiently routing traffic to the appropriate AI model based on the task at hand.-Led the integration of the ELK stack into the system to enable rapid data and error analysis, while also implementing a real-time study status update feature in the live environment. Developed a study tracking system that streamlines failure analysis, reducing resolution time by 85%.-Led the design, development, and deployment of a subscription service that provides control over the number of studies processed by AI, enabling efficient sales of AI products.-Developed an interactive web dashboard using Dash and Plotly to showcase real-time data visualizations, charts, and statistics for monitoring and analysis.-Integrated five AI models into the radiologists' live worklist, enabling the triage of positive cases from negative ones.-Developed a RAG-based LLM project using open-source Ollama models and integrated it with the live worklist to automatically proof radiologist reports, reducing proofing time by 60%.-Used Pegasus from Hugging Face to generate impressions from findings using a summarization tool, and integrated it into the live worklist for automatic generation of impressions.-Implemented a centralized storage solution for all AI model virtual machines by utilizing a Samba mount, enabling seamless data access and efficient management across multiple systems. Show less