Mudit Agrawal Email & Phone Number

Kirkland, WA, US

North Reading, MA, US

Leading Perception and Match at Amazon Robotics for robotic manipulation

Menlo Park, CA, US

Reality Labs - Because teleporting is cooler than commuting!Working on eye tracking and codec avatar disambiguation.

Redmond, Washington, US

• Led iris recognition and health sensing with a team of software engineers and research scientists. Pioneered and spearheaded the research, development and deployment using cutting-edge deep learning models like LSTMs.
• Coded the first prototype, and built the team to ship world’s first iris recognition biometrics on AR glasses featuring enterprise-grade security (with a 1 in 100,000 FAR), 3D anti-spoofing measures, and login times of.
• Implemented hybrid machine learning models, merging traditional and advanced deep learning models. Reduced optics, sensors, and MTF needs by an order of magnitude, enabling miniature cameras. Ran the system solely on.
• Pioneered innovative health solutions in AR, forging new paths for the company. Authored the inaugural white paper and spearheaded collaborations with esteemed institutions, including Stanford University, hospitals.
• Led the development of 3 health sensing solutions: brain health (Autism, Parkinson's, Alzheimer's), ocular health with eye-tracking, and cardiovascular health with PPG data. Implemented advanced deep learning models.
• Collaborated closely with the hiring team establishing streamlined workflows to expedite candidate screening and enhance overall candidate satisfaction. Conducted hundreds of interviews to successfully select.

Redmond, Washington, US

• Job Scope: Pioneered iris reco biometrics solution for HoloLens, Windows Phone and Surface Laptop
• Built the iris recognition algorithm equaling the accuracy of any Windows Hello biometrics solution.
• Developed data collection protocols and tools to optimize workflows (with QA) collecting thousands of user data under GDPR, privacy and ethical AI guidelines, generating the most diverse dataset for training, compared.
• Designed and optimized the end-to-end metrics system to do billions of image comparisons, bringing down the time of a job run from 72 to 4 hours, saving millions of dollars in cloud compute.

Redmond, Washington, US

• Job Scope: Developed machine learning solutions for millions of customers in Windows and Office.
• Developed unstructured text detection in natural scenes using LSTM for Bing streetside imagery. This work helped blur car license plates, anonymizing data for training and validation.
• Developed and deployed pen-based shape classification engine (in PowerPoint), spatial probability and language model based soft-keyboard word correction, and mouse-to-touch input mapping using geometric algorithms..

College Park, MD, US

Document Image Processing - noise removal, page segmentation, character recognition

Palo Alto, CA, US

• Job Scope: Developed machine learning solutions for pen computing interfaces for Indic scripts.
• Patented and developed hybrid gesture handwriting recognition system and pen computing interface. Created a universal XML standard (UPX format) for platform-independent representation and annotation of digital ink.
• Developed and led low-cost data collection projects, overseeing comprehensive field studies conducted nationwide, resulting in substantial cost savings for HP in data collection.

Phd, Computer Science

Masters Of Science, Computer Science

B.Tech., Computer Science