-
-
Security Policies
-
Dynamic Analysis Results
-
Security Alerts
-
Dashboard
-
-
User Access Control
-
PR review
-
Dashboard 1.2
-
Static Analysis result
-
Security Assistant
-
API Security Testing
-
Dynamic Analysis
-
Mobile App Security
-
LLM Security
-
Landing page 1.2
-
Project Management
-
Security report of project
-
Extension security report
-
Landing Page
-
Reports
TRILUX - The AI-Powered Security Command Center
Inspiration
I've spent countless hours watching MSPs struggle with an impossible mission of protecting dozens of clients across hundreds of endpoints while hackers get smarter every single day. One night, after hearing about yet another breach at a small business that simply couldn't afford enterprise security, it clicked, what if we could give every organization the same security superpowers that Fortune 500 companies have?
The Gemini 3 announcement was the catalyst. When I saw the enhanced reasoning capabilities and reduced latency, I knew this was the moment to build something that could actually think like a security expert. Not just another tool that spits out alerts, but a genuine AI security partner that understands context, learns continuously, and acts before threats become breaches.
What it does
TRILUX is your AI-powered security co-pilot that finds vulnerabilities before hackers do.
We're solving a $200B problem, cybersecurity attacks cost businesses $4.45 million on average per breach, with 43% of attacks targeting small and medium businesses who can't afford dedicated security teams.
Here's what TRILUX delivers:
🔍 Threat Modeling Engine
Gemini 3 analyzes your entire application architecture, mobile apps, web platforms, APIs, cloud infrastructure, and maps potential attack vectors. It doesn't just find vulnerabilities; it understands why they exist and prioritizes them based on real-world exploitability.
⚡ Static Code Analysis
Our AI scans source code across 20+ languages and frameworks, detecting security flaws that traditional scanners miss. We support mobile apps, browser extensions, GenAI applications, IoT firmware, and cloud-native microservices.
🎯 Dynamic Security Testing
TRILUX doesn't just scan, it attacks. Our AI-powered DAST engine simulates real hacker behavior, but with full context of your system architecture. It's like having a white-hat hacker on payroll 24/7.
🛡️ Real-time LLM Security
With AI becoming ubiquitous, we protect against prompt injection, jailbreaks, and data leakage in real-time. Our fine-tuned detection models block malicious intent before damage is done.
📋 Compliance Automation
Continuous SOC2, ISO27001, HIPAA, GDPR, and PCI-DSS compliance checks with auto-generated audit-ready reports.
Global Impact: If TRILUX prevents even 1% of SMB breaches globally, we're talking about saving millions of businesses from financial ruin and protecting billions in customer data.
How we built it
The Magic Behind TRILUX:
Gemini 3 API The brain of our operation. We use Gemini 3's enhanced reasoning for:
- Multi-step threat analysis with chain-of-thought security reasoning
- Contextual vulnerability understanding that goes beyond pattern matching
- Natural language security reports that humans actually understand
- Real-time decision-making for threat blocking with minimal latency
Google Antigravity Our secret weapon for rapid development. Antigravity's AI-powered coding assistant accelerated our prototype from concept to working product in record time.
AI Agent Architecture:
- RAG (Retrieval-Augmented Generation): Our security knowledge base includes CVE databases, threat intelligence feeds, and compliance frameworks, all queryable in real-time by Gemini 3
- MCP (Model Context Protocol): Enables seamless integration between our scanning engines and the AI reasoning layer
- Custom Fine-tuned Models: Specialized models for malicious intent detection and code vulnerability classification
Tech Stack:
- Frontend: Next.js with Tailwind CSS for a blazing-fast, beautiful dashboard
- Backend: Python/Django with PostgreSQL and Redis for robust, scalable APIs
- Security: JWT authentication with role-based access control
- Infrastructure: Google Cloud Platform (Cloud Run, Cloud SQL, Cloud Storage, Vertex AI)
- Security Tools Integration: Nuclei, custom scanning engines, API security testers
Challenges we ran into
1. Making AI "Think" Like a Hacker
Early versions of TRILUX found vulnerabilities but couldn't explain how to exploit them or why they mattered. We spent days crafting prompts and fine-tuning until Gemini 3 could reason about attack chains like an experienced pentester.
2. Balancing Speed vs. Accuracy
Real-time security monitoring can't have false positives, but it also can't be slow. We implemented a tiered analysis system: quick heuristic checks first, deep AI analysis for suspicious patterns. Latency dropped from 2 seconds to under 200ms.
3. Building the Security Knowledge Graph
We had to ingest, structure, and make queryable thousands of CVEs, attack patterns, and compliance requirements. Our RAG pipeline went through 4 major rewrites before it could handle the complexity.
4. The "Context Window Challenge"
Large codebases exceed context limits quickly. We built an intelligent chunking system that preserves semantic relationships between code modules while staying within token limits.
Accomplishments that we're proud of
✅ 30+ Security Tools in One Platform We consolidated what used to require a dozen different subscriptions into a single AI-powered dashboard
✅ 80% Reduction in Manual Security Work Our AI handles the tedious vulnerability analysis that used to take hours
✅ Real-time LLM Protection We're one of the first to offer production-ready prompt injection and jailbreak defense
✅ Sub-200ms Threat Detection Fast enough for real-time blocking without impacting user experience
✅ Compliance Reports in Seconds What used to take auditors weeks now takes our AI seconds to generate
✅ Built in Under 8 Weeks Thanks to Gemini 3 and Antigravity, we went from idea to functional product incredibly fast
What we learned
Gemini 3 is a Game-Changer for Security
The reasoning capabilities are genuinely different. Previous models could spot patterns; Gemini 3 actually understands attack logic and can explain it in human terms.
RAG is Essential for Domain Expertise
No foundation model knows every CVE. Combining Gemini 3's reasoning with our security knowledge base created something far more powerful than either alone.
The Best Security is Proactive
Every second matters. By the time a traditional scanner alerts you, damage may already be done. Real-time AI analysis changes the equation entirely.
User Experience Matters in Security
Security tools are notorious for being unusable. We focused on making TRILUX beautiful and intuitive, because a tool nobody uses protects nobody.
What's next for TRILUX
Phase 1: Q2 2026
- Public beta launch for MSPs and security teams
- Extended language and framework support
- SOC as a Service integration
Phase 2: Q3 2026
- VS Code extension for developer-first security
- CI/CD pipeline integration for shift-left security
- Mobile app for on-the-go security monitoring
Phase 3: Q4 2026
- Enterprise tier with custom AI model training
- Threat intelligence marketplace
- Bug bounty platform integration
The Vision:
Every organization from two-person startups to Fortune 500 enterprises should have access to world-class security. TRILUX makes that possible by putting Gemini 3's intelligence to work protecting what matters most.
Architecture Diagram
┌─────────────────────────────────────────────────────────────────────────────────────┐
│ USER INTERFACES │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │
│ │ Web │ │ Mobile │ │ VS Code │ │ API │ │
│ │ Dashboard │ │ App │ │ Extension │ │ Access │ │
│ │ (Next.js) │ │ │ │ │ │ │ │
│ └──────┬──────┘ └──────┬──────┘ └──────┬──────┘ └──────┬──────┘ │
└─────────┼────────────────┼────────────────┼────────────────┼────────────────────────┘
└────────────────┴────────────────┴────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────────────────────────┐
│ TRILUX AI ENGINE (Gemini 3) │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌───────────┐ │
│ │ Threat │ │ Static │ │ Dynamic │ │ Real-time │ │ Security │ │
│ │ Modeling │ │ Code │ │ Security │ │ LLM │ │ Research │ │
│ │ Engine │ │ Analysis │ │ Testing │ │ Monitoring │ │ Lab │ │
│ └─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘ └───────────┘ │
│ │
│ ┌─────────────────────────────────────────────────────────────────────────────────┐│
│ │ RAG Pipeline + Security Knowledge Graph ││
│ │ (CVE Database • Threat Intel • Compliance Frameworks • MCP) ││
│ └─────────────────────────────────────────────────────────────────────────────────┘│
└─────────────────────────────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────────────────────────┐
│ CORE SERVICES │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌───────────┐ │
│ │ Auth │ │ PostgreSQL │ │ Redis │ │ Vector │ │ Report │ │
│ │ (JWT) │ │ (Data) │ │ (Cache) │ │ Database │ │ Generator │ │
│ └─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘ └───────────┘ │
└─────────────────────────────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────────────────────────┐
│ GOOGLE CLOUD PLATFORM INFRASTRUCTURE │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌───────────┐ │
│ │ Cloud Run │ │ Cloud SQL │ │ Cloud │ │ Vertex AI │ │ IAM & │ │
│ │ (Containers)│ │ (Managed) │ │ Storage │ │ (Fine-tune) │ │ Security │ │
│ └─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘ └───────────┘ │
└─────────────────────────────────────────────────────────────────────────────────────┘
Built with ❤️ using Gemini 3, Google Antigravity, and Google Cloud Platform
TRILUX - Because every hacker's worst nightmare should be your best friend.
Built With
- agentic-ai
- antigravity
- django
- django-rest-framework
- docker
- gemini-image-generation
- gemini3
- genai
- google-cloud
- llm
- mcp
- next.js
- postgresql
- python
- rag
- redis
- vector-database


Log in or sign up for Devpost to join the conversation.