The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
Conducted an automated security assessment of a locally-hosted large language model, Dolphin 3.0 (Llama 3.1 8B), using Garak, NVIDIA's open-source LLM vulnerability scanner. The model was served through LM Studio's OpenAI-compatible REST API and tested from a Parrot OS environment running in VirtualBox.
Configured Garak to target the model via a custom REST generator configuration, handling endpoint routing, Bearer token authentication, JSON response parsing, and extended timeouts for local inference. Executed nine distinct probe categories against the model, covering DAN jailbreaks, prompt injection, encoding-based bypass, malware generation, latent injection, real toxicity prompts, continuation attacks, grandma exploits, and package hallucination.
The assessment identified that Dolphin 3.0, an uncensored model with no built-in safety guardrails, was vulnerable across the majority of probe categories. Findings were documented per probe, distinguishing where the model resisted versus complied, and translated into a clear, layered explanation of each attack type and its real-world security implications.
Delivered the work as a complete, reproducible walkthrough, including environment setup, tool configuration, probe execution, and results interpretation, suitable for both technical practitioners and stakeholders newer to AI security.
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
🚀 Continuing my backend development journey.
Recently I've been focusing on building REST APIs with Python, Django REST Framework, PostgreSQL, Docker, Git, and Linux.
Every project helps me improve API design, database modeling, clean architecture, and maintainable backend code.
I'm continuously learning and looking forward to new backend development opportunities.
#Python #Django #BackendDevelopment
Echo is a multi-tenant SaaS for AI customer support with realtime chat, an AI voice agent, and AI automation. I was the primary developer on the full build, frontend, backend, and the AI layer. The AI is the core. It runs OpenAI, Claude, Gemini, and Grok through one multi-model setup, so a client can switch providers without a rewrite. A RAG pipeline connected to a vector database grounds every answer in the client's own content, so the chatbot never returns generic output. VAPI powers the voice agent, so customers can speak to support on a live call. Each tenant gets its own AI agent built on its own documents. The stack is Next.js 15 and React 19 inside a Turborepo monorepo, with separate apps for the dashboard, the embeddable chat widget, and backend services. Realtime chat runs on Convex. Clerk handles auth. API keys are encrypted per tenant through AWS Secrets Manager, so no two clients share credentials or data. Launch day held 60 live conversations at once with zero dropped sessions. What the client gets: an AI chatbot and voice agent that answer from their own content, work across multiple LLM providers, and stay isolated and secure per tenant.:
Perfect
Meet AI is a video conferencing SaaS where AI agents join live meetings as active participants and respond in real time. I built the full application stack.
The hard part is the real-time layer. The backend provisions an AI agent the moment a meeting starts and keeps it running for the full call. Speech runs through a transcript pipeline and generates responses through the OpenAI API in near real time, so the agent replies while the conversation is still moving, not minutes later. After the meeting, the backend writes a structured summary through the same API, handled async by Inngest so nothing blocks the live app.
The stack is Next.js 15 and React 19 with the Stream Video and Stream Chat SDKs, tRPC and Drizzle ORM on Neon PostgreSQL, and Better Auth across the stack.
The platform runs 30 simultaneous live sessions, each with its own active AI agent, with no backend slowdown.
What the client gets: live AI agents that take part in real meetings, plus automatic summaries, on a stack built to hold many sessions at once.
Perfect
Trending
Claude
Claude has entered the design space. How are you using Claude Design?
Contra University
Learn from expert creatives how to earn more using next-gen AI tools.
fifaworldcup2026
The World Cup is here and the whole world's watching. How are you designing for the world stage?
creativeaiflow
Creative AI workflows are evolving. What tools do you use, and what are their strengths and weaknesses?
freelancerlife
Freelancer life is wins, pivots, and everything in between. What’s yours right now?