> For the complete documentation index, see [llms.txt](https://hasithz.gitbook.io/hasith-abayakoon/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://hasithz.gitbook.io/hasith-abayakoon/selected-work/projects/project-1.md).

# EagleEye

Applied AI for camera-based security analysis, led from concept to deployment

**Research Lead, May 2023 – May 2024.** Applied AI product. Status: deployed; the product is live at [eagleeyes.ai](https://eagleeyes.ai/).

## Problem

Organisations already own large numbers of security cameras, but footage is only as useful as the people watching it. Continuous human monitoring does not scale, and incidents involving weapons or suspicious behaviour are easy to miss across a wall of feeds. EagleEye applies AI to existing camera infrastructure so that potential threats, suspicious activity, and weapons are surfaced automatically rather than discovered after the fact.

## My role

I was Research Lead and led the flagship AI project from concept to deployment. I designed the core AI systems, choosing the model families and the processing architecture, with scalability, reliability, and performance as the design goals.

## Constraints

* **Existing cameras, not purpose-built sensors.** The system had to work with whatever feeds customers already had, with their varying resolutions, angles, and lighting.
* **Many concurrent streams.** Analysis has to keep pace with live video across sites, so throughput and resilience mattered as much as model accuracy.
* **A product, not a prototype.** Models had to be packaged, deployed, and kept running rather than evaluated once in a notebook.

## Decisions

* **Model families.** Vision transformers and graph neural networks, implemented in PyTorch, for recognition on camera frames.
* **Asynchronous processing.** Kafka for messaging between services and Celery for background task execution, so that ingestion, inference, and alerting were decoupled.
* **Containerised deployment.** Docker for packaging the models and services, so the same build ran in development and in deployment.

## Evidence

* Product site: [eagleeyes.ai](https://eagleeyes.ai/)

## Outcome

The project went from concept to a deployed product during my tenure. No accuracy or operational figures are published here.

[Back to selected work](/hasith-abayakoon/selected-work/projects.md)