Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,69 @@
+---
+license: mit
+language:
+- en
+task_categories:
+- text-generation
+- question-answering
+tags:
+- interconnection
+- solar
+- utilities
+- coordination
+- fine-tuning
+- qwen
+pretty_name: "IX Coordination Tasks"
+size_categories:
+- 10K<n<100K
+---
+
+# IX Coordination Tasks
+
+Instruction-tuning dataset for interconnection coordination in utility-scale solar and storage projects. Built from real Salesforce task logs (76K+ entries across 5,700+ projects) to train models on utility-specific workflows and rejection resolution.
+
+## Dataset Summary
+
+- **35,005** instruction-response pairs (ShareGPT/ChatML format)
+- **31,504** training / **3,501** validation (90/10 split)
+- Covers **37+ utilities** including Duke Energy, PG&E, Xcel, Consumers Energy, ComEd, AEP, PPL, and more
+- Organized into 4 task categories
+
+### Categories
+
+| Category | Examples | Description |
+|---|---|---|
+| **Rejection → Resolution** | 2,852 | Utility rejects interconnection application (system size, design, missing docs) → resolution steps |
+| **Missing Item → Collection** | 6,112 | Missing signed documents, HOI, FIN, utility bills, placards → collection and submission procedures |
+| **Utility Communication** | 4,647 | Calling/following up with utilities → what to ask, what to have ready |
+| **Follow-Up Workflow** | 21,394 | Routine status checks, portal verification, next-action determination |
+
+## Data Format
+
+Each example follows the ChatML/sharegpt format with system/user/assistant roles:
+
+```json
+{
+  "messages": [
+    {"role": "system", "content": "You are an expert interconnection coordination assistant..."},
+    {"role": "user", "content": "Project: 1606MERE\nUtility: Duke Energy Progress NC\n..."},
+    {"role": "assistant", "content": "Rejection Analysis for 1606MERE | Duke Energy Progress NC\n..."}
+  ]
+}
+```
+
+## Intended Use
+
+Fine-tuning small-to-medium LLMs (3B-8B) with QLoRA/LoRA for interconnection coordination assistance. The model learns to:
+
+1. Identify the correct utility-specific workflow
+2. Diagnose rejection causes from task log context
+3. Provide concrete resolution steps with follow-up timelines
+4. Reference utility portal requirements and submission procedures
+
+## Source Data
+
+Built from the Interconnection Dash repository's Salesforce task log exports:
+- `All-Task-Logs.csv` (76,092 rows)
+- `All-Projects-All-Time.csv` (utility name and state enrichment)
+
+Retrieved May 5, 2026.