DevFest Fresno - Build with AI
Details
Join us for DevFest Fresno, a unique afternoon dedicated to the latest in AI technology. Part of the worldwide DevFest season, focusing on building with AI and exploring its innovative applications. Engage with fellow tech enthusiasts on October 18th, Saturday, at Fresno State University's Grosse Industrial Tech IT 101 auditorium.
The best parking place is the green P20 parking lot North of campus across Barstow Avenue just north of the Grosse Industrial Tech building complex. After crossing Barstow Avenue, the IT 101 is the large auditorium at the Westernmost side of the building.
Agenda:
12:30-1pm Warming up and setting up
1pm - 2pm Nandan Gupta and Nayan Goel - Principal Application Security Engineer @ Fintech
Jailbreak-as-a-Service: Automating Prompt Injection with PromptInjector
Prompt injection is no longer theoretical — it's happening in production systems and AI-integrated products right now. As developers rush to integrate LLMs into everything from chatbots to medical software, security teams are struggling to test them with the same rigor we apply to traditional apps. Enter PromptInjector: a new open-source tool that brings sqlmap-style automation to LLM testing.
In this talk, we’ll introduce PromptInjector, a purpose-built framework for identifying prompt injection vulnerabilities using both static and AI-generated adaptive attacks. Designed for red and blue teams alike, it comes with over 1,000 curated prompts, automatic behavioral analysis, campaign tracking, and a modular config system.
We’ll demonstrate how PromptInjector can:
Simulate and detect 7 classes of prompt injection, including jailbreaks, system prompt leaks, and role manipulation
Automatically adapt to LLM responses and evolve its attack strategy (like a fuzzing engine for prompts)
Generate detailed reports with risk scoring, success rates, and confidence analysis
Help teams test AI systems defensively and prevent unsafe outputs before they hit end users
Expect a live demo, walkthrough of the architecture, real-world testing examples on open models and APIs, and a breakdown of what worked (and what broke). We'll also share insights on LLM behavior during red team simulations and how attackers are evolving.
Takeaway: If you're shipping LLMs or securing them, you need to know how prompt injection works — and how to break your own models before others do.
2pm - 3pm Laiylaly Mandujano - Linguist
Building AI Agents using Vertex AI Studio
In this hands-on workshop, we'll be going over the basics of creating an agent using Vertex AI studio. We'll create a travel agent that uses rules and gen AI to help users plan a trip.
3pm - 4pm Kiran Kumar Reddy Pamuru - Google
AI-Powered Row-Level Security: Gemini-Enhanced Data Protection for Modern Apps
This presentation explores implementing intelligent row-level security using Google's Gemini AI and modern cloud architectures. As enterprises increasingly rely on AI-driven applications, traditional static security models fall short of protecting dynamic, context-aware data access patterns. Our case study demonstrates how integrating Gemini's natural language processing capabilities with Firebase Security Rules and Cloud SQL creates adaptive security layers that respond intelligently to user behavior and access patterns.
The implementation leverages Gemini API to analyze access requests in real-time, using natural language understanding to interpret complex business rules and automatically generate appropriate security predicates. Firebase Studio Code Assist accelerated development by generating boilerplate security functions, while Gemini CLI streamlined deployment workflows across multiple environments. The AI-enhanced system dynamically adjusts permissions based on contextual factors including user department, project involvement, and temporal access patterns, moving beyond rigid role-based access control.
Results demonstrate significant security improvements: unauthorized access attempts decreased by 75% within the first quarter, while false positive security blocks dropped by 60% due to Gemini's contextual understanding. The AI system correctly interpreted 95% of complex access scenarios that previously required manual intervention. Development velocity increased substantially, with Firebase Studio Code Assist reducing security rule implementation time from weeks to days. The Gemini-powered audit system automatically generates compliance reports, reducing manual effort while improving accuracy.
This session will provide practical insights for developers implementing AI-enhanced security architectures, demonstrating how Google's AI tools can transform enterprise data protection while maintaining developer productivity and operational efficiency.
4pm - 5pm Kiran Kumar Reddy Pamuru - Google
GenAI Compensation Intelligence: Building AI-Driven Payroll Systems with Google Tools
This presentation showcases building next-generation compensation management systems using generative AI and Google's developer ecosystem. Modern enterprises managing complex compensation structures face unprecedented challenges in accuracy, fairness, and adaptability. Our framework demonstrates how Gemini's reasoning capabilities, combined with Firebase's real-time database and AI-powered development tools, creates intelligent compensation systems that understand nuanced business rules and adapt to organizational changes.
The architecture integrates Gemini API for natural language processing of compensation policies, automatically translating complex HR documents into executable business logic. Firebase Studio Code Assist significantly accelerated development, generating sophisticated calculation functions and validation routines. Jules coding agent helped architect the microservices infrastructure, while Gemini CLI automated deployment pipelines across development, staging, and production environments. The system processes millions of compensation calculations with AI-enhanced accuracy validation and anomaly detection.
Gemini's language understanding capabilities enable the system to interpret complex compensation rules written in plain English, automatically generating corresponding calculation logic. The AI continuously learns from compensation patterns, identifying potential errors before they impact payroll processing. Real-time Firebase integration ensures instant synchronization across all compensation modules, while Gemini-powered analytics provide executives with natural language insights into compensation trends and optimization opportunities.
Implementation results show remarkable improvements: calculation accuracy increased to 99.9%, policy change implementation time reduced from weeks to hours, and employee compensation disputes decreased by 85%. The AI system correctly interprets 98% of complex multi-variable compensation scenarios. This session will demonstrate practical implementation strategies for developers building AI-enhanced financial systems using Google's generative AI ecosystem.
5pm - 5:30pm Raffles and Closing
Agenda
---
Hosted By
Csaba Toth, GDG lead, WTM ambassador
Generative AI Engineer, GDG lead, WTM ambassador, tech meetup enthusiast.
Grace Ann Aranico, Organizer + WTM ambassador
Ashley Rice,
Complete your event RSVP here: https://gdg.community.dev/events/details/google-gdg-fresno-presents-devfest-fresno-build-with-ai/.