Rudblest/projedanismanai-v2-qwen3-14b
Rudblest/projedanismanai-v2-qwen3-14b is a 14 billion parameter, Turkish-language instruction-tuned model based on Qwen3-14B. Fine-tuned with 3,043 Turkish examples, it specializes in assisting TEKNOFEST and TÜBİTAK competition participants with project development, report writing, strategy, and error correction. This model offers enhanced performance over its predecessor by leveraging a more powerful base model and optimized training, specifically designed for project consultancy in Turkish.
Loading preview...
ProjeDanışmanAI v2: Turkish AI Consultant for TEKNOFEST & TÜBİTAK
ProjeDanışmanAI v2 is a specialized 14 billion parameter language model, fine-tuned by Rudblest on the Qwen3-14B base model. It is designed to act as an AI consultant for participants in TEKNOFEST and TÜBİTAK competitions, providing guidance throughout their project development process in Turkish.
Key Capabilities
- Report Writing: Assists with KTR/PTR technical report sections and provides examples.
- Project Development: Helps generate new project ideas and outlines roadmaps.
- Summary & Analysis: Summarizes project documents and analyzes strengths/weaknesses.
- Strategy: Offers jury-focused presentation strategies and strengthens originality sections.
- Error Correction: Identifies deficiencies in existing report sections.
- Contextual Filtering: Politely declines questions outside the scope of TEKNOFEST/TÜBİTAK topics.
What makes it different?
This v2 iteration significantly improves upon its predecessor (v1, based on Mistral-Nemo-12B) by utilizing the more powerful Qwen3-14B, which offers performance comparable to Qwen2.5-32B while maintaining a 14B VRAM footprint. It was trained using QLoRA 4-bit with Unsloth on 3,043 Turkish instruction-output pairs, with optimized parameters to prevent overfitting. A crucial operational note is the requirement to prepend /no_think to system prompts, as the model was trained with its thinking mode disabled to prevent empty <think> tags in responses.
Good for
- TEKNOFEST and TÜBİTAK competition participants needing assistance with project documentation and strategy.
- Developers seeking a specialized Turkish-language AI for technical project consultancy.