Edge AI
Computer Vision
FY25–26
Realtime Industrial Anomaly Detection
A YOLO-based edge AI pipeline built for Bar & Rod Mill video analytics — detects
equipment anomalies, near-miss events and safety violations from live plant camera
feeds, with low-latency inference optimised for on-prem edge hardware via OpenVINO.
Forms the applied core of my M.Tech research at IIT Bhilai.
YOLOv8
OpenVINO
Edge Deployment
Industrial IoT
Anomaly Detection
- Role
- Researcher / Edge AI Developer
- Domain
- Bar & Rod Mill industrial video analytics
- Model
- YOLO-based detection / pose-oriented pipeline
- Runtime
- OpenVINO / edge inference
- Hardware
- Jetson / industrial edge server
- Output
- Anomaly events, confidence, timestamp, frame evidence
- Status
- M.Tech research prototype / deployment-focused case study
Multi-Modal Alert Summarization
An LLM/RAG system that ingests sensor telemetry, vision-model outputs and plant log
streams, then produces plain-language operator alerts — shrinking the gap between an
anomaly firing and a human understanding what it means.
LangChain
RAG
Multi-modal
LLMs
Autograder+
A grading framework that pairs static code analysis — linting, complexity, test
coverage — with semantic LLM feedback, giving students structured, rubric-aligned
comments instead of a single pass/fail signal.
Python
Static Analysis
LLMs
Smart Weed Remover Robot
An NVIDIA Jetson Nano based vision system paired with a 6-DOF parallel robot to
identify weed candidates and perform targeted mechanical removal in crop rows.
NVIDIA Jetson Nano
Computer Vision
6-DOF Parallel Robot
AgriTech
Zeroshot Text-to-Image AI
A research notebook exploring zero-shot text-to-image generation — prompt-conditioning
strategies, diffusion sampling parameters and qualitative evaluation across creative
and technical prompt sets.
Diffusion Models
Prompt Engineering
Jupyter
Edge AI
Computer Vision
2021
Computer Pointer Controller
A gaze-controlled mouse pointer built on a four-model OpenVINO pipeline — face
detection, head-pose estimation, facial-landmark detection and gaze estimation —
chained together for real-time CPU inference.
OpenVINO
Gaze Estimation
Computer Vision
Smart Queuing System
Edge-deployment job scripts that benchmark a people-counting model across
manufacturing, retail and transportation hardware profiles on Intel DevCloud —
comparing CPU, GPU, FPGA and VPU performance trade-offs.
OpenVINO
Edge Benchmarking
Intel DevCloud
Healthcare AI
Computer Vision
2020
Peter Moss COVID-19 AI Research (xDNN)
Open-source contribution implementing an eXplainable Deep Neural Network (xDNN) for
real-time diagnostic-support predictions — designed to run at the edge while keeping
its decision process interpretable for clinicians.
xDNN
Explainable AI
Edge Inference
Edge AI
Computer Vision
2021
People Counter App Edge
An OpenVINO Toolkit application that detects and counts people in a video stream,
tracks dwell time, and pushes live totals to an MQTT broker and Node.js dashboard for
facilities analytics.
OpenVINO
MQTT
Computer Vision