wincode/kerala-crime-detective-gemma

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:1BQuant:BF16Ctx Length:32kPublished:Apr 2, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

The wincode/kerala-crime-detective-gemma is a 1 billion parameter Gemma 3-based causal language model fine-tuned by wincode. It specializes in understanding and responding to Kerala crime reports, FIR details, and cyber fraud cases in Malayalam, English, and Manglish. This model uniquely combines serious investigative analysis with comedic and cultural references, offering distinct modes for humorous, professional, or cybercrime-focused responses.

Loading preview...

Kerala Crime Detective - Multilingual AI for Crime Solving

wincode/kerala-crime-detective-gemma is a fine-tuned Gemma 3 1B model designed to process and respond to crime reports in Malayalam, English, and Manglish, incorporating Kerala-specific cultural nuances and humor. Developed by wincode, this model offers a unique blend of serious investigative capabilities and comedic elements, making it distinct from general-purpose LLMs.

Key Capabilities

  • Multilingual Crime Analysis: Understands and responds to crime reports in Malayalam, English, and Manglish, including local slang and cultural references.
  • Diverse Response Modes: Features distinct modes for:
    • Malayalam Comedy: Solves crimes with humor, local jokes, and Manglish.
    • Serious English: Provides professional CID-style investigation, including evidence, suspects, and legal sections.
    • Cyber Crime Expert: Specializes in common cyber frauds like UPI fraud, SIM swap, sextortion, and fake job scams, offering immediate victim steps and legal recourse.
    • Mixed Style: Combines comedic and serious advice.
  • Specialized Training Data: Fine-tuned on the "Kerala Crime Comedy Dataset," which includes various crime categories from gold theft to cyber fraud and even light-hearted cases.

Good For

  • Simulating Crime Investigations: Ideal for educational or entertainment applications requiring crime-solving scenarios with a regional flavor.
  • Multilingual Content Generation: Generating responses that blend English, Malayalam, and Manglish for specific crime-related queries.
  • Exploring AI with Cultural Context: Demonstrating how LLMs can be adapted to specific cultural and linguistic contexts beyond standard English tasks.

Limitations

  • Trained on a relatively small dataset (21+ examples), which may affect response consistency.
  • Not a substitute for real police or legal advice; intended for educational and entertainment purposes only.

Finetunes
1 models