Sumit Pardhiya

Product Manager

Hyderabad, Telangana, India3 yrs 11 mos experience

AI ML PractitionerAI Enabled

Key Highlights

Expert in real-time computer vision and deep learning systems.
Proven track record in multilingual OCR and barcode recognition.
Strong MLOps experience for scalable AI model deployment.

Stackforce AI infers this person is a Computer Vision and Machine Learning expert with a focus on embedded systems.

Contact

Skills

Core Skills

Optical Character Recognition (ocr)Barcode RecognitionMachine LearningObject DetectionGesture RecognitionMlopsAnomaly Detection

Other Skills

Multilingual AI SystemsPaddleOCRTesseractPARSeqZXingC++KubernetesLinuxYOLOv7OCROpenCVMLYOLOXKubeflowGitOps

About

Image Processing Engineer at ACG World, building real-time Computer Vision and Deep Learning systems for automated visual inspection — turning noisy, real-world imagery into reliable production decisions. I specialize in object detection, OCR, and image segmentation, with a focus on making models run fast and accurately on embedded/edge hardware. At ACG I work on multilingual OCR, barcode recognition, and medicine segmentation. Before that, at Samsung R&D Institute India, I built and deployed on-device AI for TV platforms — object detection, content recognition, gesture tracking — and led MLOps for reproducible model development. Day-to-day tools: Python, C++, PyTorch, TensorFlow, OpenCV, ONNX, YOLO, and an MLOps stack of Git, DVC, and MLflow. Outside engineering, I write — technical pieces on Medium and GeeksforGeeks about ML and computer vision, plus travel writing documenting the places I explore. I think the best ideas come from curiosity, whether that's a new architecture or a new city. Open to connecting with people working on computer vision, edge AI, and applied ML — feel free to reach out.

Experience

3 yrs 11 mos

Total Experience

3 yrs 2 mos

Average Tenure

9 mos

Current Experience

Acg world

Image Processing Engineer

Sep 2025 – Present · 9 mos · Hyderabad · On-site

🔹 Multilingual OCR System (PaddleOCR + PARSeq + Tesseract)
Designed and implemented a multilingual OCR pipeline supporting Chinese, English, and Russian text recognition. Fine-tuned PaddleOCR for Chinese datasets, improving recognition accuracy across diverse fonts and low-quality images. Integrated PARSeq for English text and optimized Tesseract for Russian OCR, enhancing overall system robustness and language coverage.
🔹 Barcode Detection & Decoding (ZXing - C++)
Implemented barcode decoding using ZXing in C++, optimizing inference performance and accuracy. Conducted performance benchmarking (latency, throughput, warm-up vs steady-state) and improved decoding efficiency through preprocessing and format-specific optimizations.

Optical Character Recognition (OCR)Multilingual AI Systems

Samsung r&d institute india

2 roles

Machine Learning Engineer

Jul 2022 – Sep 2025 · 3 yrs 2 mos · Delhi, India

🔹 UI Perceiver (YOLOv7 + OCR)
Designed and deployed an on-device YOLOv7 model for TV screen analysis, achieving 98% accuracy. Integrated OCR for text-based UI element detection, improving text recognition accuracy by 25%. Optimized inference to <50ms, enhancing real-time performance on embedded devices by 40%.
🔹 Hand Gesture Detection (OpenCV + ML)
Led development of a gesture tracking system using advanced computer vision techniques. Improved detection accuracy from 60% to 85% through iterative model tuning, testing, and algorithm optimization.
🔹 App Exploration Graph (YOLOX + Visual Similarity)
Built a frame comparison module to cluster visually similar frames, achieving 97% classification accuracy. Developed a YOLOX-based object detection model for focus detection with 89% accuracy, streamlining UI automation.
🔹 MLOps Automation (Kubeflow + GitOps + DVC)
Created an end-to-end MLOps pipeline supporting scalable development and deployment. Improved experiment tracking by 30% using Kubeflow, accelerated deployment by 40% with GitOps, and enhanced model versioning with DVC, reducing iteration time by 25%.

KubernetesLinuxMachine Learning

Machine Learning Intern

Jan 2022 – Jul 2022 · 6 mos · Delhi, India

🔹 Abnormal Screen Detection (Smart TVs)
Led the end-to-end design and development of a deep learning model to detect abnormal or corrupted screens on Smart TVs. Achieved 97% accuracy in anomaly detection, significantly enhancing automated quality assurance and device reliability.

LinuxConvolutional Neural Networks (CNN)Anomaly Detection