Computer Vision & Deep Learning Engineer - VISTRAS TECHNOLOGIES
(2025-10)
Video Analytics, Industrial Inspection Backend & Vision Pipelines
- Developed a high-availability, real-time video analytics platform using deep learning frameworks, achieving a 98% detection accuracy threshold for mission-critical object identification and industrial classification.
- Engineered a hardware-integrated inference system synchronized with PLC (Programmable Logic Controller) hardware triggers to capture and process images on-demand for automated inspection workflows.
- Optimized the complete end-to-end vision pipeline—from camera frame ingestion via RTSP streams to GPU-accelerated execution—guaranteeing sub-20ms latency thresholds critical for high-speed production.
- Implemented spatial intelligence logic including ROI (Region of Interest) filtering and spatial bounding box matrices to handle precise, zone-specific object counting.
Computer Vision & Deep Learning Engineer - VISTRAS TECHNOLOGIES
(2025-10)
Autonomous Automated Annotation Application ("Auto-Labeler")
- Architected and optimized an automated zero-intervention image and video annotation system leveraging Hugging Face Foundation Models to streamline large-scale dataset preparation.
- Engineered the pipeline to simultaneously process and label 10+ distinct object classes, slashing data collection and validation lifecycle overhead by 96%.
- Trained, fine-tuned, and validated multiple YOLO architectures (v8, v11n, v11s) utilizing a hybrid curation strategy of manual verification and automated preprocessing to establish rigorous ground truth data.
- Elevated core object detection performance from 80% to 93% accuracy through structured feature engineering and iterative hyperparameter tuning.
Computer Vision & Deep Learning Engineer - VISTRAS TECHNOLOGIES
(2025-10)
Thermal Imaging & Environmental Anomaly Detection Suite
- Engineered a real-time analytical monitoring system using computer vision to continuously analyze thermal video streams, map temperature deviations, and isolate critical threshold breaches.
- Programmed custom optics compensation logic, ambient temperature offset configurations, and adaptive environmental calibration loops to guarantee high sensing accuracy across diverse physical spaces.
- Designed and deployed an automated event-triggered alerting pipeline integrated with webhooks to instantly dispatch metadata and anomaly timestamps to remote operational units for proactive risk mitigation.
