Python
Engineering
Computer Vision Engineer
Umang Shah
I build intelligent computer vision systems that move from research to real-world deployment, with deep learning, VLMs, and real-time CV pipelines.
Current role
AI R&D Engineer
Core domain
Real-time CV systems
Specialty
Multimodal AI Systems
Delivery style
Research to production
Available
<>
{ }
</>
( )
About
Research mindset, engineering execution
I build computer vision systems that turn research ideas into deployable AI products.
I am a Computer Vision Engineer focused on production AI pipelines, deep learning, multimodal models, and modern vision frameworks. My work connects model accuracy, system performance, and practical business impact.
My current work involves YOLO-based detection, homography-driven camera alignment, GroundedSAM, OCR, vision language models, heatmaps, ROI automation, background synthesis, and scene-level LLM summaries for spatial intelligence and security monitoring.
I also bring applied research experience in facial recognition, visual attention, CNN and MobileNet models, ONNX deployment, and large-scale identity systems. I enjoy translating complex vision problems into scalable AI systems that deliver measurable results.
Primary stack
Python, PyTorch, OpenCV
Model work
YOLO, GroundedSAM, ONNX
AI systems
VLMs, GraphRAG, OCR
Current focus
Computer vision, ML systems, and applied AI
Technologies I work with
My Skills
A practical mix of computer vision, deep learning, backend engineering, and deployment tools used to build production AI systems.
Python
Engineering
Computer Vision
Core ML
YOLO
Detection
OpenCV
Vision
PyTorch
Deep Learning
Keras
Deep Learning
scikit-learn
ML
GroundedSAM
Applied AI
VLMs
Multimodal AI
ONNX
Deployment
Django REST
Backend
Flask
Backend
MongoDB
Database
GraphRAG
LLM Systems
OCR
Vision
GitHub Actions
CI
Experience
Experience across research and production
A concise view of the roles behind my computer vision, applied ML, and software engineering work.
BluB0X Security
AI Research and Development Engineer
Building real-world vision intelligence systems for security and facility operations. The work combines YOLO, GroundedSAM, OCR, homography, heatmaps, and LLM workflows to turn camera data into useful spatial insights.
July 2024 - PresentAndover, MA
Visual Attention Laboratory, UMass Boston
Research Assistant
Worked on facial recognition and visual attention research with a focus on model reliability, image quality, and deployment readiness. This included CNN and MobileNet-based models, ONNX conversion, and automatic annotation workflows.
May 2023 - May 2024Boston, MA
Atsign Inc.
Software Engineer Intern
Helped improve the Python developer experience for atProtocol applications by contributing SDK functionality, tests, compatibility checks, examples, and documentation.
May 2023 - Jul 2023Remote
GlobalVox
Associate Software Engineer
Built computer vision and backend systems across facial identification, emotion recognition, number plate detection, and enterprise platforms. This role shaped the foundation for combining ML models with scalable APIs.
Jul 2020 - Jun 2022Ahmedabad, India
GlobalVox
Artificial Intelligence Intern
Started with practical applied ML and computer vision work using Python, scikit-learn, Pandas, NumPy, and image-based model experiments before moving into a full-time engineering role.
Jan 2020 - Jul 2020Ahmedabad, India
Education
Formal training in CS and AI
The academic side of the portfolio spans computer science, artificial intelligence, machine learning, and production-oriented software fundamentals.
Sep 2022 - May 2024
UMass Boston
Master's degree, Computer Science
Grade: 4.0/4.0
Specialized in computer vision, neural networks, machine learning, algorithms, and compilers with applied work in facial recognition and object detection.
Boston, MA
2018 - 2020
Gujarat University
MSc, Artificial Intelligence
Grade: 4.0/4.0
Built a foundation in machine learning, deep learning, advanced Python, image-based tasks, model training, evaluation, and optimization.
Ahmedabad, India
2015 - 2018
AES Institute of Computer Studies
BCA, Computer Programming
Grade: 3.97/4.0
Studied data structures and algorithms, object-oriented programming, software development, and database management systems.
Ahmedabad, India
Projects
Selected work from the live portfolio
These projects show the mix of computer vision, research, and practical software systems behind the Machine Learning Engineer focus.
Selected project
Person Re-Identification
A research project focused on matching people across different cameras and viewpoints in surveillance settings. The challenge is handling lighting changes, camera angles, occlusion, and other real-world variation.
SurveillanceMatchingResearch
Selected project
Banana Classification
Detects and classifies bananas as naturally or artificially ripened using data collected from markets and farms. Built to show practical computer vision applied to a focused real-world problem.
OpenCVClassificationCV
Selected project
Initial Centroids for K-Means
A research effort aimed at improving K-Means clustering by choosing better initial cluster centers and estimating the number of clusters more effectively.
ClusteringOptimizationResearch
Selected project
Government Prosecutors' Portal
A portal for public prosecutors to manage cases, calendars, case studies, notifications, and elections, with an admin panel and forum features for collaboration.
PortalAdminForum
Selected project
Notes Management System
A Google Keep-style note app with reminders, doodles, file attachments, sharing, and email reminders, backed by MySQL for storage and retrieval.
NotesReminderSharing
Selected project
File Transfer Tool
A desktop file-sharing tool that transfers files over UDP and HTTPS without requiring internet access or a flash drive.
UDPHTTPSOffline
Contact
Let's talk
For roles, research, or collaborations, email is the fastest way to reach me.
Get in touch
Have a project, role, or idea?
If the problem is messy, visual, or real-time, that is usually the interesting part. Send it over.
Email me
Weird constraints are welcome. The best projects usually start there.