Devyani Marathe

Developed a computer vision model (human pose estimation) based on real-time video data, TensorFlow / Pytorch / OpenCV, trained and tested on more than 1k images with accuracy of over 90%. Optimized inference pipelines for accuracy and speed and reduced the latency by 35% achieving 20+ FPS on CPU systems and improved the consistency of detections in difficult environments.

AI Intern at Techsaksham (Microsoft & SAP) (2024-12 – 2025-01)

• Developed a computer vision model (human pose estimation) based on real-time video data, TensorFlow / Pytorch / OpenCV, trained and tested on more than 1k images with accuracy of over 90%.
• Optimized inference pipelines for accuracy and speed and reduced the latency by 35% achieving 20+ FPS on CPU systems and improved the consistency of detections in difficult environments.

Software Developer Intern at Edunet Foundation (2024-07 – 2024-08)

• Developed NLP-built chatbot workflows using python IBM watsonX and LLM pipelines with 85% intent accuracy.
• Integrated IBM Cloud APIs into conversation workflows to cut response time by 25%.
• Collaborated with CI/CD and cloud services to build and automate experimentation with API-based ML services via IBM AutoAI to decrease the time for model development by 30-40%.

B.E. in IT – NBN Sinhgad Technical Institute Campus, Pune (2022-11 – 2026-06)