Add POC stub

1_Pooling/config.json

@@ -0,0 +1,9 @@
+{
+  "word_embedding_dimension": 384,
+  "pooling_mode_cls_token": false,
+  "pooling_mode_mean_tokens": true,
+  "pooling_mode_max_tokens": false,
+  "pooling_mode_mean_sqrt_len_tokens": false,
+  "pooling_mode_weightedmean_tokens": false,
+  "pooling_mode_lasttoken": false
+}

README.md

@@ -0,0 +1,62 @@
+---
+language: en
+license: apache-2.0
+tags:
+  - sentence-transformers
+  - feature-extraction
+  - embeddings
+  - semantic-search
+  - contrastive-learning
+library_name: sentence-transformers
+pipeline_tag: feature-extraction
+---
+
+# org-shared-embeddings
+
+Shared embedding model for cross-team semantic search over documentation,
+runbooks, and resolved ticket summaries.
+
+Hosted under the **organization** namespace for centralized inference endpoint billing.
+
+## Model description
+
+| Property | Value |
+|---|---|
+| Base model | `sentence-transformers/all-MiniLM-L6-v2` |
+| Output dimension | 384 |
+| Pooling | mean |
+| Normalization | L2 |
+| Max sequence length | 256 |
+
+## Intended use
+
+- Internal doc search (`/v1/search/docs`)
+- Duplicate ticket detection
+- Clustering for QA review sampling
+
+## Usage
+
+```python
+from sentence_transformers import SentenceTransformer
+
+model = SentenceTransformer("matt-ts/org-shared-embeddings")
+query = "How do I rotate API keys for the staging environment?"
+doc = "Staging key rotation: open IAM console, select service account..."
+similarity = model.similarity(query, doc)
+print(similarity)  # tensor([[0.72]])
+```
+
+## Deployment
+
+| Endpoint | Region | Instance |
+|---|---|---|
+| `embeddings-prod-us` | us-east-1 | `gpu-l4-small` |
+| `embeddings-prod-eu` | eu-west-1 | `gpu-l4-small` |
+
+## Version history
+
+| Version | Date | Notes |
+|---|---|---|
+| v1.2.0 | 2026-01-08 | Added runbook corpus (+12k docs) |
+| v1.1.0 | 2025-09-22 | Ticket summary fine-tune |
+| v1.0.0 | 2025-06-01 | Initial MiniLM baseline |

config.json

@@ -0,0 +1,22 @@
+{
+  "architectures": ["BertModel"],
+  "model_type": "bert",
+  "torch_dtype": "float32",
+  "transformers_version": "4.38.2",
+  "vocab_size": 30522,
+  "hidden_size": 384,
+  "num_hidden_layers": 6,
+  "num_attention_heads": 12,
+  "intermediate_size": 1536,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "attention_probs_dropout_prob": 0.1,
+  "max_position_embeddings": 512,
+  "type_vocab_size": 2,
+  "initializer_range": 0.02,
+  "layer_norm_eps": 1e-12,
+  "pad_token_id": 0,
+  "embedding_dimension": 384,
+  "pooling_mode": "mean",
+  "normalize_embeddings": true
+}

modules.json

@@ -0,0 +1,14 @@
+[
+  {
+    "idx": 0,
+    "name": "0",
+    "path": "",
+    "type": "sentence_transformers.models.Transformer"
+  },
+  {
+    "idx": 1,
+    "name": "1",
+    "path": "1_Pooling",
+    "type": "sentence_transformers.models.Pooling"
+  }
+]

tokenizer_config.json

@@ -0,0 +1,10 @@
+{
+  "do_lower_case": true,
+  "model_max_length": 256,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "cls_token": "[CLS]",
+  "unk_token": "[UNK]",
+  "mask_token": "[MASK]",
+  "tokenizer_class": "BertTokenizer"
+}