Hugging Face's logo
Activity Feed

AI & ML interests

None defined yet.

Recent Activity
thaki-AI  updated a collection about 1 hour ago
EmbeddingGemma 2 Edge — Quantization
thaki-AI  updated a collection about 1 hour ago
EmbeddingGemma 2 Edge — Quantization
thaki-AI  updated a collection about 1 hour ago
SkillRet — Agent Skill Retrieval
View all activity

ThakiCloud 's collections 17

EmbeddingGemma 2 Edge — Quantization
Quantized EmbeddingGemma 2 text towers that query an existing BF16 index without re-embedding the corpus.
ChainCheck — Chain-Sensitive Evidence Judges
Evidence-sufficiency judges for RAG (4B/9B/27B) tuned with ChainCheck controls, plus the kit. Gains are on real entities, not synthetic ones.
Merchant Desk — Korean Small-Business Customer Desk
동네 음식점 고객 문의를 사장님 대신 받는 정책 모델. 의도·슬롯·행동을 JSON으로 고르고, 사실은 매장 프로필에서만 가져온다.
Qwen3.8-27B — Serving Quantization
Serving-quantization suite for Qwen3.8-27B, all built from the same base so you can pick by hardware generation.
Human-KO — Korean Style-Aligned 27B
Full-weight checkpoints aligning Qwen3.8-27B to Korean conversational style: style, CJK suppression, safety DPO, serving quantization.
PACE — Enterprise Policy Decisions: Learned or Reasoned
Which of seven actions should an enterprise agent take under a written policy? Generation is removable; test-time reasoning is not.
Qwen3-Coder-30B-A3B — Serving Quantization
Quantization suite for the coding MoE 30B-A3B. Four formats plus an expert-pruned variant.
SkillRet — Agent Skill Retrieval
Retrieval embeddings, benchmark, and index that let an LLM agent find the right skill inside a library of 1,000+ skills.
EmbeddingGemma 2 Edge — Quantization
Quantized EmbeddingGemma 2 text towers that query an existing BF16 index without re-embedding the corpus.
ChainCheck — Chain-Sensitive Evidence Judges
Evidence-sufficiency judges for RAG (4B/9B/27B) tuned with ChainCheck controls, plus the kit. Gains are on real entities, not synthetic ones.
PACE — Enterprise Policy Decisions: Learned or Reasoned
Which of seven actions should an enterprise agent take under a written policy? Generation is removable; test-time reasoning is not.
Merchant Desk — Korean Small-Business Customer Desk
동네 음식점 고객 문의를 사장님 대신 받는 정책 모델. 의도·슬롯·행동을 JSON으로 고르고, 사실은 매장 프로필에서만 가져온다.
Qwen3-Coder-30B-A3B — Serving Quantization
Quantization suite for the coding MoE 30B-A3B. Four formats plus an expert-pruned variant.
Qwen3.8-27B — Serving Quantization
Serving-quantization suite for Qwen3.8-27B, all built from the same base so you can pick by hardware generation.
SkillRet — Agent Skill Retrieval
Retrieval embeddings, benchmark, and index that let an LLM agent find the right skill inside a library of 1,000+ skills.
Human-KO — Korean Style-Aligned 27B
Full-weight checkpoints aligning Qwen3.8-27B to Korean conversational style: style, CJK suppression, safety DPO, serving quantization.