Chaukidar

Vercel
application badge
Created by team PunkRockers on July 12, 2026
AMD Developer CloudAMD ROCmCodex
Unicorn Track

Chaukidar helps teams evaluate whether AI models remain safe beyond English-only testing. It audits LLM responses across South Asian languages and harm categories such as fraud, cyber abuse, self-harm, hate, and violence. The platform compares English seed prompts, translation-baseline prompts, and native-adapted prompts to reveal where model safety behavior changes across language, culture, and phrasing. Built for AMD Hackathon Track 3, Chaukidar uses AMD ROCm/vLLM notebook results for compute-backed audits, supports Fireworks model comparisons, re-judges imported outputs with a multilingual judge, and generates dashboards, risk heatmaps, track comparisons, and audit reports.

Category tags:
GithubPresentationDemo
Explore more applications
Streamlit
application badge

router_007_v3

router_007_v2 is a Track 1 agent that records **zero billable tokens**: every answer is computed inside the container by a Qwen2.5-7B-Instruct model bundled in the Docker image and served in-process with llama-cpp-python

roc_auc_half

Claude CodeChatGPTAMD ROCm

Lexyprep

Agent knows how to make thorough legal research based on my experience of winning legal cases. It understands your issue and advises on procedure

Lexyprep - Do you have a case

AMD Developer CloudAMD ROCmCodexGemma

video-captioning-agent

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Mauris at cursus dolor. Phasellus porta, sapien ut molestie aliquet, sapien velit.

memovate

AMD Developer Cloud

career launch ai

CareerLaunch AI: An intelligent workspace architect that synthesizes personalized technical roadmaps, project directories, and AMD-optimized hardware configurations using Llama 3 AI via Fireworks.

bloatware slayer

ChatGPTGemini 3 FlashLlama 3.1

TokenRouter - Efficient AI Task Routing Agent

A lightweight routing agent that solves simple tasks (math, sentiment) locally for zero token cost, and only escalates harder tasks to an LLM via the Fireworks AI API — minimizing token usage while maintaining accuracy.

Train&Try

ChatGPTAnthropic Claude