.jpeg)
Hey 👋; My name is Dev Patel and I'm a Machine Learning Developer Specializing in Deep Learning and Bioninformatics. Previous experience includes project management, blockchain engineer, and software development.
| From 🇨🇦 in Toronto, Ontario 🍁 | 🗓 June 28 | devpatelio.netlify.app | tks.life/profile/dev.patel | github.com/devpatelio |
<aside> 🌘 STATUS: N/A
</aside>
OTHER: | bitclout.com/u/devpatelio | newsletter; https://bit.ly/3xR7Hhe | hello-devpatel.medium.com/ | linkedin.com/in/dev-patel-25456219a/ | calendly.com/devpatel-scheduling/one-on-one |
ML Open-Source Frameworks and Languages: Tensorflow, PyTorch, Octave, Cuda + Python, C++
Packages and Platforms: TKinter, Numpy, Pandas, NLTK, SpaCY, GloVe, Seaborn, Matplotlib, Django, Flask, OpenCV, Azure, VertexAI, PyTorch Lightning, Kaggle
Web App FullStack: Django, HTML/CSS, Vanilla JS, SQLite, Heroku
Expertise: Computer Vision, Genetic Data, Data Structures and Algorithms, Supervised Machine Learning, Unsupervised Deep Learning, Artificial Intelligence, Data and Signal Preprocessing, APIs, Scripting and Terminal, Version Control Git, Bioninformatics, Cloud Deployment for ML, Blockchain Tooling, etc.
Other: Outreach, Solution Synthesis, and Pitch Decks, Consulting and Proposal Development, Project Management and Leadership, Research, Agile-Lite Production Work Flow
DOKTOR I → Diagnostics Tool for Patients with Range of Serverity for Diabetic Retinopathy Using Deep Learning and Convolutional Neural Networks [Nov 2020-Jan-2021]
Developed a deep learning computer vision algorithm that rivaled the test and train accuracy of traditional ophthalmologists by 35.4% on national average. The system is built using preprocessed training data of patients with several levels of diabetic retinopathy in severity (No-DR to Proliferate-DR) and accurately diagnose the level of severity early on. It can also make specific predictions for patients with little to no DR to collect more long-term snapshots of diabetic patients. Used a resenet with elements of transfer learning, trained on the GPU → Validation accuracy of 80.45% → clinical diagnostics aim to alleviate much of the burdgen with early diagnostics.