Hackathon Portal
AI Tinkerers - Dublin
Sign in

AI Tinkerers Dublin: Agentic Architect Hackathon with Lyzr AI

Sign in
Engineering Lead at Stealth AI Startup - Stealth AI Startup
Own AI-native architecture and backend pipelines remotely with no equity gimmicks.
All Jobs
Team

AdGuard AI

Project Concept

AdGuard AI — Autonomous Brand Ad Safety Screening

AdGuard AI is an agentic brand safety screening system that automatically
evaluates publisher websites before ad campaigns go live — protecting brands
from appearing next to harmful, inappropriate, or brand-damaging content.

Problem

Brands lose millions in reputation damage when ads appear next to violent,
hateful, or inappropriate content. Manual screening is slow, expensive, and
doesn’t scale across thousands of publisher URLs.

Solution

AdGuard AI uses a multi-agent architecture built on Lyzr Studio to
autonomously screen publisher URLs against brand-specific safety rules using
the GARM (Global Alliance for Responsible Media) framework.

How It Works

  1. URL Researcher — analyses website content and context
  2. Safety Classifier — scores content across 11 GARM categories
  3. Brand RAG Agent — applies brand-specific rules from a knowledge base
  4. Results Reporter — outputs scores, decisions and sends alerts via Slack and Gmail

Tech Stack

  • Lyzr Studio — multi-agent orchestration
  • Claude Opus 4 — manager agent reasoning
  • Agentic RAG — brand safety knowledge base (Disney, LEGO, Nike, McDonald’s, Coca-Cola)
  • GARM Framework — industry standard brand safety classification
  • Slack + Gmail — real-time campaign alerts

Output

Every URL gets a safety score (0-100), GARM flag, and decision: Run / Review / Block.

Entry

Status: Not Started

Team Roster

You must be registered for the event to view the team message board.

Steven Gerard Mascarenhas Team Lead RSVP Approved

Student at University of limerick
Steven Gerard Mascarenhas is an Aspiring AI/ Machine Learning Engineer at HackIreland, with approx 3 years of experience. He holds an MSc in Artificial Intelligence & Machine Learning from the University of Limerick and a BE in Computer Engineering from Don Bosco College of Engineering. Steven is a Software Developer with an AI focus, having built a skin lesion classification model for cancer detection achieving over 90% accuracy. He automated the pipeline using AWS services including S3, Lambda, and SageMaker. Steven is open to work and seeking knowledge sharing, co-founders, speaking opportunities, and sponsorships. He is open to introductions and prefers contact via email.
End-to-end AI and ML deployments, Vision Transformers, Computer Vision, Retrieval Augmented Generation (RAG), AWS SageMaker, Lambda, S3, Terraform, CI/CD, Technical architecture, Product review, Knowledge sharing, Co-founding (Technical and Business), Speaking opportunities, Sponsorships, Software Engineering, AI/ML research.
I built a skin lesion classification for skin cancer detection which achieved over 90% accuracy using a pretrained hugging face vision transformer for low cost on a production environment .automated the pipeline using aws services like s3 ,lamda for preprocessing and sagemaker for deployment and inference. I also developed ai powered medical chatbot using retreival augmented generation and deployed on aws by using terraform and github ci/cd actions