Antony Oduor@oduortoni
1
Events attended
1
Submissions made
Kenya
3+ years of experience
About me
Fullstack Software Developer with hands-on experience in system design, development, and optimization. Skilled in Go, PHP, C, and JavaScript, with notable framework experience in Laravel. Passionate about turning client needs into functional software solutions.
🤝 Top Collaborators
.png&w=640&q=75)
🤓 Latest Submissions
PromptGuard AI
PromptGuard AI is an enterprise agent firewall for LLM apps. It sits between the users, models and agent tools and then scores prompts, model outputs, uploaded text and tool calls before allowing the action to proceed. As organisations rapidly deploy AI agents that can write code, pull data from databases, send emails, and change infrastructure, the attack surface increases exponentially. A single prompt injection can transform a helpful assistant into a data exfiltration tool. A leaked API key in a model output can put entire cloud environments at risk. Tool calls without guards can wipe production servers. PromptGuard AI addresses this challenge by providing a centralised security checkpoint, an AI-era firewall, to inspect each interaction between humans, models and tools in real-time. The system employs a hybrid detection strategy that combines 35 deterministic regex rules (executing in less than 1 ms) with Google Gemini AI semantic analysis (to capture new and obfuscated threats that patterns alone cannot catch). This two-tier architecture offers both speed and depth: known attack patterns are blocked instantly, while the AI layer catches subtle social engineering, encoded payloads, and adversarial prompt techniques. All decisions are explainable and auditable, and security teams can override them.
19 May 2026
