← Back to Projects
This project is a sub-project of AgCloud
Security
Mentored by: Vast Data
Security - Cloud-based platform for agricultural data management and analytics
Apache Kafka
Apache Flink
Microservices
Docker
Prometheus
Grafana
Airflow
PostgreSQL
PostGIS
MinIO
FastAPI
CI/CD
pytest
Python
PyQt6
PostgreSQL
MQTT
Description
Security sub-project of AgCloud. A comprehensive cloud platform for managing agricultural operations, data, and analytics. Provides centralized storage, processing, and visualization of farm data including crop monitoring, weather integration, equipment management, and predictive analytics. Features include multi-tenant architecture, real-time dashboards, and API integrations.
Team Members
Cohort: Data Science Bootcamp 2025 (Data)
Responsibilities:
Designed and implemented a scalable real-time streaming system for security event detection from distributed surveillance camera feeds.
Trained and benchmarked YOLO-CLS and MobileNetV3 image classification models for masked-person detection, comparing accuracy and runtime performance and integrating the chosen model into the production pipeline.
Researched and evaluated multiple approaches for climbing-activity detection, selecting and integrating the most accurate solution into the system.
Selected and integrated an animal classification model to enable species-based detection of unauthorized intrusions.
Designed and developed a domain-specific language (DSL) enabling free-text search over security events, translating user queries into structured, executable filters.
Implemented a full security observability UI layer, encompassing real-time alert streaming, historical event querying with detailed inspection, and analytical dashboards for system monitoring and evaluation.
Designed and implemented a centralized Alert Management system with real-time geo-spatial alert visualization based on GPS locations and a unified notifications center.
...and more contributions not listed here
Responsibilities:
Added Prometheus metrics to Python microservices using prometheus-client, exposed /metrics, and built Grafana dashboards for latency and error-rate monitoring, later expanded to sensor-health visibility.
Research : Built a simulated edge security-rover and an ingestion pipeline that publishes imagery and telemetry via MQTT, stores images in MinIO, sends metadata to Kafka, and enforces consistent Pydantic schemas and path conventions.
Delivered end-to-end fence-breach detection: trained and evaluated YOLOv8 models, selected YOLOv8n (conf=0.35), exported to ONNX, and deployed a FastAPI inference service that pulls images from MinIO, publishes alerts to Kafka/Alert-Manager, and writes results to PostgreSQL — improving false-positive rates while maintaining high recall.
Authored a model-comparison report supporting architecture decisions.
Integrated detection flow into the streaming pipeline: MinIO bucket notifications → Kafka → Flink → inference service → database + alerts.
...and more contributions not listed here