Mohit Gaikwad
Building production-grade computer vision systems for surveillance, industrial inspection, and real-time analytics.
About Me
Production-focused Computer Vision Engineer with 3.5+ years of experience designing and deploying real-time AI systems across video analytics, surveillance, industrial inspection, and robotics. Specialized in end-to-end vision pipelines, from model optimization using TensorRT, Triton and DeepStream to edge deployment on Jetson devices. At Wobot.ai, built multi-camera ReID systems covering 500+ live cameras across 20+ restaurant locations. At Griffyn Robotech, currently engineering industrial inspection pipelines processing 200 units per minute with 95%+ accuracy. My work sits at the intersection of computer vision and systems engineering, writing code that ships, stays stable, and solves real operational problems.
Professional Timeline
A journey of scaling vision systems from embedded prototypes to multi-camera production deployments.
- Sentinel: 100 RTSP stream traffic platform (DeepStream + Redis + Kafka) with face recognition, ANPR, vehicle classification, occupancy detection and Traffic Signal Violations
- High-speed bottle inspection: 200 bottles/min, 95%+ accuracy, <1.5% false acceptance rate. 90% manual effort reduction. Multi camera Inspection System
- Wrist-watch defect detection: 73% cycle-time reduction (30s → 8s)
- Supervised and Unsupervised anomaly detection.
- Drive-thru analytics: Multi camera vehicle ReID across 500+ cameras, 20+ QSR locations
- Reduced identity mismatch from 12–15% → 2–5%
- Staff-assisted checkout detection system (>20 checkouts/hour)
- Zero-shot object detection & classification
- Automated training + evaluation pipelines (Docker, CI/CD, MongoDB, MLOps)
- Multi-object tracking: BoTSort, ByteTrack, Hybrid Sort
- Smart traffic monitoring pipeline for edge devices
- ANPR, object detection, detection → tracking → analytics workflow
- Python, MySQL, Raspberry Pi
- Autonomous factory automation prototype (AI + Computer Vision)
- Jetson Nano, Jetson Xavier NX, Raspberry Pi
- Introduction to embedded AI deployment
Featured Projects
Sentinel: Traffic Monitoring
100 RTSP streams · 30 FPS · DeepStream + Triton + Redis + Kafka. Detection → ANPR → Face Recognition → Traffic Signal Violations → Occupancy Analytics across multiple camera views.
Multi-Camera Drive-Thru ReID
500+ cameras · 20+ locations · Identity mismatch reduced from 12–15% → 2–5% using custom cross-camera ReID with iterative failure-case retraining.
High-Speed Bottle Inspection
200 bottles/min · >95% accuracy · <1.5% false acceptance · ~90% manual effort reduced. Unsupervised + supervised hybrid anomaly detection at sub-200ms latency.
Wrist-Watch Defect Detection
73% inspection cycle-time reduction (30s → 8s/unit). Dual-camera synchronized capture providing 360° surface coverage with automated defect classification across multiple surface types.
OneAnomaly
Hybrid cloud-edge anomaly detection software. FAISS vector search · Local GPU inference · Cloud UI.
OneProductIQ
AI-powered E-commerce platform using LLMs, VLMs, RAG, and agentic tool calling for multimodal product search and recommendations.
OneEye
An advanced computer vision system prototype exploring scalable, real-time monitoring architectures with a focus on edge deployment and multi-stream processing.
Let's Build Something Together
Available for full-time roles, contract projects, and technical collaborations in computer vision and AI systems engineering.