Headline

Computer Vision Engineer · Detection, Segmentation & Tracking · Edge AI (TensorRT/ONNX) · Traffic & Smart City Systems · Open to Opportunities

About

I build computer vision systems that work in the real world - not just in research notebooks. Over 2+ years at Metafusion, joining early as part of the founding team, I helped shape and ship the company's core CV product stack from scratch: intelligent traffic monitoring, video incident detection, multi-lane free-flow tolling, and ANPR pipelines deployed across live infrastructure. My work spans the full modern vision stack. On the modelling side: object detection (YOLO, DETR), instance segmentation (SAM, Mask R-CNN), multi-object tracking (ByteTrack, SORT), and pose estimation. On the deployment side: TensorRT and ONNX optimization for edge hardware, C++ OpenCV pipelines for latency-critical systems, and Docker/Kubernetes for production delivery. I've also integrated Vision-Language Models into production environments using vLLM and Ollama - connecting multimodal reasoning into existing CV pipelines where classical detection alone wasn't enough. I care about systems that are fast, reliable, and actually deployed - not demo-ware. If you're building something in computer vision, edge AI, or video analytics and want to talk, reach me at tejuskavishwar78@gmail.com

Location

Gurugram

Work experience

Computer Vision Research Engineer at Metafusion, Jun 2024 - Present · 2 yrs 1 mo, Noida, Uttar Pradesh, India, • Architected and shipped three production video analytics suites - ITMS (traffic management), VIDES (incident detection), and MLFF (free-flow tolling) - by building end-to-end pipelines combining YOLO/DETR detection, ByteTrack/SORT multi-object tracking, and ANPR with OCR, resulting in live deployments processing concurrent camera streams across smart city and highway infrastructure. • Optimized detection and segmentation models (YOLO, Mask R-CNN, SAM) for edge hardware deployment using TensorRT and ONNX quantization, and maintained C++ OpenCV pipelines for latency-critical scenarios where Python overhead was not acceptable. • Extended production CV pipelines with Vision-Language Model (VLM) integration via vLLM and Ollama, enabling multimodal scene reasoning in scenarios where classical detection alone was insufficient - deployed using Docker and Kubernetes.

Computer Vision Intern at Metafusion, Oct 2023 - May 2024 · 8 mos, India

Data Science Intern at Innomatics Research Labs, Feb 2023 - Jun 2023 · 5 mos, Hyderabad, Telangana, India, • Built and evaluated supervised ML models (classification and regression) using Scikit-learn and Pandas, improving baseline performance by 20% through feature engineering and hyperparameter tuning. • Conducted EDA across 10+ structured datasets, surfacing statistical patterns that informed model selection and preprocessing decisions.

Computer Vision Intern at Awiros, Jul 2022 - Aug 2022 · 2 mos, Gurugram, Haryana, India, • Implemented real-time face detection and recognition pipeline using OpenCV and deep learning, integrated into live surveillance video streams for production testing. • Built a pose estimation module for a fitness application (push-up counter), gaining hands-on experience with human keypoint detection and temporal consistency on video.

Education

Bachelor of Technology - BTech, CSE (AI and ML), Maharshi Dayanand University (MDU), Rohtak