Exploring the Frontiers of AI, Space, and Technology
I'm Kathan Mehta, a Machine Learning expert and proud alumnus of Carnegie Mellon University. By day, I serve as the AI Lead at Satelytics, architecting and deploying cutting-edge AI for geospatial data analysis. I'm passionate about leveraging technology to create practical, impactful solutions and am always eager to connect with fellow innovators.
Professional Experience
Satelytics
AI Lead
Dec 2022 - Present
- Leading end-to-end strategy for AI projects, overseeing planning, implementation, deployment, and continuous optimization of geospatial data processing workflows.
- Driving a 30% improvement in model efficiency through advanced optimization techniques, enabling rapid data ingestion, processing, and delivery with minimal human intervention.
- Mentoring a team of AI engineers, fostering a culture of innovation and technical excellence.
AI Engineer
Dec 2020 - Nov 2022
- Developed and optimized preprocessing workflows for diverse satellite imagery, including monochromatic, multispectral, and hyperspectral data from platforms like Pleiades, WorldView, Sentinel, and PlanetScope.
- Pioneered the development and automation of key geospatial algorithms for Methane Detection, Leak Detection, and identification of Invasive Grasses, improving detection accuracy by over 25%.
- Engineered stereo DSM/DTM workflows for advanced tree height detection, health analysis, and comprehensive vegetation management solutions.
Carnegie Mellon University
Graduate Research Assistant - Biomedical Image Guidance Lab
Jan 2020 - Nov 2020
- Led a research project on needle tip localization in Ultrasound images, applying both classical computer vision and novel Deep Learning approaches to track needle paths and speckle patterns.
- Contributed to the paper "Ultrasound-Based Tracking Of Partially In-Plane, Curved Needles," published in the 2021 IEEE International Symposium on Biomedical Imaging (ISBI).
- Developed a Deep Learning algorithm for detecting signs of Diabetic Retinopathy in retinal fundus images, contributing to early-stage disease identification research.
Education
Carnegie Mellon University
Master of Science - MS, Electrical & Computer Engineering
2019 - 2020
University of the Cumberlands
Master of Business Administration - MBA
2023 - 2024
Dwarkadas J. Sanghvi College of Engineering
Bachelor of Engineering - BE, Electrical, Electronic and Communications Engineering
2014 - 2018
Featured Projects
Image Colorization using CNNs
Developed a CNN-based deep model combining low-level local features and high-level global features to colorize grayscale images, utilizing a U-Net style encoder-decoder architecture.
View on GitHub →Speech to Text Translation
Designed a Listen-Attend-Spell model using a pyramidal BiLSTM encoder and attention-based LSTM decoder, achieving a 11.57 test Levenshtein distance on the WSJ dataset.
View on GitHub →Face Classification and Verification
Built a CNN using MobileNetV2 for efficient face classification. Employed transfer learning to use embeddings for face verification, achieving 93.3% AUC on the test dataset.
View on GitHub →Blockchain-based Voting System
Designed a secure voting system using a hybrid of Proof-of-Work for identity authentication and Proof-of-Authority for vote storage, secured with AES and RSA encryption.
View on GitHub →Distributed File System
Built a global file system with REST APIs for client operations, maintaining a hierarchical view at the naming server and ensuring data durability through replication.
View on GitHub →Braille Display for the Visually Impaired
Designed an award-winning (1st Prize, DJ Spark) Braille display system using OCR to mitigate communication barriers for the visually impaired.
View on GitHub →Publications
Ultrasound-based Tracking of Partially In-plane, Curved Needles
2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI)
Read Publication →Technical Skills
Languages & Frameworks
Tools & Platforms
Beyond the Code
When I'm not building AI models, here's what I'm passionate about.