Smit Patel — AI/ML & Computer Vision Engineer

About Me

Available for freelance AI/ML & Computer Vision work

🎓 I'm currently pursuing my M.E. in Artificial Intelligence & Data Science at GTU School of Engineering & Technology (2025–2027), on top of a B.Tech in Computer Science & Engineering from Ganpat University. Alongside my studies, I teach AI, Machine Learning and Deep Learning to undergraduate engineering students as a Teaching Assistant and former Lecturer — so I explain models as clearly as I build them.

🔬 My core focus is deep learning for computer vision — image classification, object detection, and applied CNN/transfer-learning pipelines (ResNet, EfficientNet, DenseNet, CNN ensembles). My peer-reviewed research applies these techniques to medical imaging, and my project work spans plant disease detection, face-mask detection, and multi-class image classification on Kaggle-style datasets.

💻 As a Software Developer at Rosix Technology I've shipped production systems end-to-end: Flask APIs integrating 15+ third-party services for 1,000+ users, a DenseNet121-based plant disease classifier, and voice-recognition features using Hidden Markov Models — plus earlier work at IBM building a Python/PostgreSQL/Docker IPR filing system.

🤝 If you need a computer vision model built, an existing pipeline debugged and pushed to higher accuracy, or a research idea turned into a working prototype, I'd love to hear about your project. Let's connect and build something awesome.

Career

• Teaching Compiler Design (Semester 6) and Computer Vision (Semester 7) to undergraduate engineering students.

• Taught Artificial Intelligence, Machine Learning, and Deep Learning to undergraduate engineering students (sem 5 & 7, BE - AI&DS, IT).
• Delivered Fundamentals of AI, Applied ML, and Deep Learning courses with practical case studies and industry relevance.
• Mentored students in SIH Hackathon projects (2025), research initiatives, and AI-based web apps.
• Completed the "AI for Educators-2025" national faculty development program by Swayam Plus (Ministry of Education) & Intel India.

UG Level Subjects Taught:
• Fundamentals of Artificial Intelligence (3154202) — Sem 5, BE AI&DS
• Introduction to Machine Learning (4350702) — Sem 5, Diploma Computer Engineering
• Deep Learning Principles and Practices (3174201) — Sem 7, BE AI&DS
• Applied Machine Learning (3171617) — Sem 7, BE IT
• Artificial Intelligence (BE04043011) — Sem 4, BE AI&DS
(GTU-affiliated curriculum)

• Developed a chatbot using Flask, integrating 15+ APIs, deployed on PythonAnywhere and Render, serving 1,000+ users.
• Built an image classification model using DenseNet121 for plant disease detection with 91% accuracy.
• Resolved multiple bugs, improving chatbot performance by 15% through iteration and testing.
• Implemented voice recognition with Hidden Markov Models, improving response time by 20%.
• Built 5 cognitive tests (e.g. clock drawing test), achieving 95% accuracy in shape detection and processing ~100 submissions/day via AWS.
• Scraped data using BeautifulSoup and Selenium, reducing extraction time by 70%.
• Designed databases and APIs for cognitive tests supporting 500+ daily users, boosting sprint completion by 15% in JIRA.

• Conducted sessions to improve fluency, pronunciation, and IELTS exam strategies.
• Designed custom lesson plans for different proficiency levels.

• Created a Python-based IPR filing system using PostgreSQL, Docker, AWS, and Jenkins, enhancing efficiency and accuracy.
• Developed a Flask application with SIFT and SURF algorithms to facilitate IPR filing for the grassroots community.
• Explored ways to visualize GitHub collaboration in a classroom setting.

• Explored methods to detect objects in live webcam and CCTV feeds.
• Developed a high-precision Face Mask Detection system using Python, CNN, TensorFlow, and Keras.
• Contributed 500+ lines of code to an established system via Git.
• Implemented real-time face capture and detection via CCTV and webcam, processing over 1,000 frames per second.
• Created an 8-page presentation and delivered multiple work-related presentations.
• Presented virtually at the World Conference on Computational Intelligence.

STTP & FDP

Developed understanding of research ecosystems, impactful research, publication strategies, and research gap identification.

Trained in Generative AI techniques and academic applications.

Acquired knowledge of literature review, research methodology, publication process, and identifying research gaps.

Gained exposure to AI-integrated pedagogy, digital learning tools, and AI-assisted teaching methodologies.

Education

Currently pursuing my M.E. in AI & Data Science, CGPA 8.33/10, deepening my focus on deep learning, computer vision, and applied research.

Graduated with a 7.10 CGPA, building a strong foundation in software systems, algorithms, and data engineering while earning certifications from several international universities and organizations.

Specialization delivered through Ganpat University's collaboration with IBM, covering Big Data & Analytics tooling, programming languages, and IBM Cloud under industry faculty from IBM.

Completed Andrew Ng's Machine Learning course, building a rigorous foundation in ML theory and practice. Won the Coursera financial-aid scholarship to take the course.

Certifications

C++ Training · C Training · Advanced C++ Training · BOSS Linux Training

Hadoop 101 · Deep Learning Fundamentals · Machine Learning with Python · SQL and Relational Databases 101 · DeepLearning.TV ML0115EN: Deep Learning Fundamentals

AWS Machine Learning Foundation · Cloud Foundation

Data Structures · Advanced Algorithms and Complexity · Machine Learning: Classification · Capstone: Retrieving, Processing, and Visualizing Data with Python · Introduction to Data Analytics for Business · Data Science Math Skills · Computer Vision Basics · Algorithmic Toolbox · Managing Big Data with MySQL · Mathematics for Machine Learning: Multivariate Calculus

PY0101EN: Python Basics for Data Science · DS0105EN: Data Science Tools

Google Cloud IAM and Networking for AWS Professionals · Managing Machine Learning Projects with Google Cloud

United States Institute of Peace — Introduction to Peace Building Micro Course · Microsoft AI Classroom Series · Corporate Finance Institute — Excel Crash Course: Spreadsheet Formulas for Finance

Skills

Computer Vision & Deep Learning

Image Classification	Object Detection	Transfer Learning
CNNs (ResNet, DenseNet, EfficientNet)	Ensemble Modeling	Medical Imaging AI
OpenCV	PyTorch / TensorFlow / Keras	Model Evaluation & Tuning

Key Skills

Data Visualization	Predictive Analysis	Statistical Modeling
Clustering	Classification	Quantitative Analysis
Data Analytics	Data Mining	Model Development
Web Scraping	ML Algorithms	Model Deployment

Technical Skills

Tools	Packages	Statistics/Machine Learning
Python	Scikit-Learn	Statistical Analysis
PostgreSQL / MongoDB / MySQL	NumPy / Pandas	Linear & Logistic Regression
AWS (EC2, S3, SageMaker)	SciPy	Clustering (KNN, K-means)
Docker / Jenkins / Git	NLTK / spaCy	Classification
Hadoop / Hive / Spark	BeautifulSoup / Scrapy	Ensemble Methods (RF, XGBoost)
Django / Flask / React / Node	Matplotlib / Seaborn / Plotly	ARIMA / SARIMA
Jupyter / Colab / R-Studio	Statsmodels	EDA
IBM Watson / Cognos / DB2	TensorFlow / Keras	Hypothesis Testing

Research

A Deep Learning-Driven Computer Vision Framework for Automated Detection, Classification, and Prognostic Analysis of Complex Medical Imaging Data

Smit Patel, Mahe Jabeen, Dr. Dhwani Modi, Indrasish Das, Dr. Subashini, Sandipan Roy · Pain, Joints, Spine, 2026;16(4):558-566 · DOI: 10.1922/pjs.16.45.202.746

A baseline deep-learning-based computer vision framework was developed and validated using the Knee Osteoarthritis Severity Grading Dataset (KneeXrayKL224), applying binary and five-class severity classification across Kellgren-Lawrence grades 0-4. The two-class task achieved 70.4% accuracy (ROC-AUC 0.70), while the five-class task reached 32.1% accuracy and an F1-score of 0.679, with an advanced convolutional architecture (image-level ensemble training, clinical data integration, and external validation identified as next steps for clinical-ready deployment).

Read the Paper DOI Google Scholar ResearchGate ORCID

Alongside this published research, I've been extending the same deep-learning-for-medical-imaging approach to broader computer vision problems — most recently a multi-class sea-animal image classification pipeline built around a fully rewritten notebook with a two-phase training strategy and a ResNet101 + EfficientNetV2M ensemble, developed for a Kaggle-style benchmark. If your project needs rigorous, publication-quality model development, this is the kind of process I bring to it.

Live AI Demo

A real, working computer vision model — not a mockup. Upload or drop a photo below and a neural network (MobileNet, trained on 1,000+ object categories) will classify it in your browser in real time.

Drag & drop an image here, or click to choose one

Or try a sample:

Powered by TensorFlow.js + MobileNet — the model runs 100% client-side in your browser. No images are uploaded to any server, and nothing leaves your device.

Achievements

Rank 3, Hackathon AI-For-India Event — built a Face Recognition Application using Python. (certificate)
Scored 97% in "Big Data Analytics: Opportunities, Challenges and the Future" by Sebastian Binnewies, School of ICT, Griffith University. (certificate)
Participated in Hacktoberfest 2022 open-source contributions, making 5 contributions across webapp projects — improving UI/design and adding features.
Rank 1, HackerRank BruteForce 2.0 Competition — successfully captured the flag.
Top 4.2% on LeetCode — solved 300+ problems in 100 days (2023), handle: jayambe36.
Rank 1, ICT Ganpat University on GeeksForGeeks, handle: jayambe36.
Won the Coursera financial-aid scholarship (2021) for the Machine Learning course by Stanford University.
Solved 1,700+ chess puzzles on chess.com.
Deep dive into Space Science (Math & Theory) at ISRO — Satellite-Based Navigation: A Journey from GPS to Mobile Phone Platform. (certificate)
Poster presentation during a hackathon — University Student Grade Prediction using Decision Tree Algorithm.
Certified Network Security Specialist (ICSI CNSS), International CyberSecurity Institute, United Kingdom — deep dive into Cyber Security. (credential)

Badges

Coding Profiles

∗ Rewrote the training notebook end-to-end around a two-phase training strategy (frozen backbone warm-up, then full fine-tuning).
∗ Built an ensemble of ResNet101 and EfficientNetV2M for multi-class sea-animal classification.
∗ Delivered comprehensive evaluation outputs (accuracy, confusion matrices, per-class metrics) to debug and maximize leaderboard accuracy.

∗ Designed and validated a deep-learning computer vision framework for automated knee osteoarthritis detection and Kellgren-Lawrence grading.
∗ Benchmarked binary and five-class classification tasks on the KneeXrayKL224 dataset, achieving 70.4% accuracy (ROC-AUC 0.70) on the binary task.
∗ Published in Pain, Joints, Spine (2026) — see the Research section for the full citation and link.

∗ Developed an image classification model using DenseNet121 for plant disease detection, reaching 91% accuracy.
∗ Integrated the model into a production Flask chatbot serving 1,000+ users.

∗ Developed a real-time Face Mask Detection system using Python, CNN, TensorFlow, and Keras.
∗ Integrated CCTV and webcam functionality, processing over 1,000 frames per second.
∗ Conducted data analytics on 10,000+ captured faces to improve detection precision.

∗ Led development of a high-accuracy (99.045%) predictive model for company bankruptcy using a financial dataset of 97 columns and 6,820 rows across 10 years.
∗ Implemented advanced data cleaning and feature selection, ensuring robust data quality and meaningful financial insight.
∗ Conducted extensive data analysis and model validation, improving reliability of financial risk assessments.

∗ Led a 3-member team building a predictive model in SPSS Modeler and Python to evaluate candidate retention.
∗ Addressed data-quality and accuracy challenges through cleaning, normalization, and reclassification, reaching 98% accuracy on the test set.

∗ Engineered an IPR (Intellectual Property Rights) filing system in Python with PostgreSQL, improving filing accuracy and operational efficiency.
∗ Used Docker for consistent deployment across environments, AWS EC2/S3 for hosting and storage, and Jenkins for CI/CD.

∗ Executed a text analytics project using the Bag-of-Words model for text visualization and NLP on 50,000+ text entries.
∗ Used POS tagging to improve text categorization accuracy by 20% and automated word/tag extraction workflows.

∗ Analyzed 9 years of champagne sales data to forecast future trends using ARIMA and SARIMA models.
∗ Performed stationarity checks with rolling statistics and the Augmented Dickey-Fuller test.

∗ Engineered a full-stack ed-tech platform (MongoDB, ExpressJS, ReactJS, NodeJS) for creating, consuming, and rating educational content.
∗ Built secure authentication, course management, and Cloudinary media integration; deployed via Vercel and Render.

I'm Smit Patel.

About Me

Career

L J University

Apollo Institute of Engineering and Technology (Affiliated to GTU)

Rosix Technology (view certificate)

EEC (English Education Centre)

IBM - Career Education Program (view certificate)

Panache Software

STTP & FDP

STTP: Exploring Research Opportunities and Impactful Research (view certificate)

FDP: Generative AI and Its Applications (view certificate)

AI for Research Excellence: From Literature Review to Publication (view certificate)

AI for Educators-2025 (view certificate)

Education

Master of Engineering, Artificial Intelligence and Data Science

Bachelor of Technology, Computer Science and Engineering (view degree)

Big Data and Analytics (view certificate)

AWS Academy Graduate - Machine Learning Foundation

Machine Learning (view certificate)

Certifications

Spoken Tutorial (IIT Bombay)

IBM Developer Skills Network

AWS Graduate Academy

Coursera

edX (IBM)

Google Cloud

Other Certifications

Skills

Computer Vision & Deep Learning

Key Skills

Technical Skills

Research

A Deep Learning-Driven Computer Vision Framework for Automated Detection, Classification, and Prognostic Analysis of Complex Medical Imaging Data

Live AI Demo

Achievements

Badges

Coding Profiles

Selected Work

Kaggle - Applied Computer Vision

Peer-Reviewed Research

Rosix Technology

Panache Software Project (Intern - Machine Learning Engineer)

IBM - Career Education Program

IBM - Career Education Program

IBM - Capstone Project

University Semester Project

University Semester Project

Industrial Project

Say Hello

Have a computer vision or deep learning project in mind? I'm available for freelance and consulting work — from proof-of-concept models to production-ready pipelines. Let's turn that idea into a working product :)

Email

Phone/Whatsapp