Pinned Loading

1. qwen3-financial-sft qwen3-financial-sft Public

   TRL + PEFT LoRA/QLoRA supervised fine-tuning of Qwen3-4B-Instruct for financial numeric QA (TAT-QA/FinQA-style) over hybrid table+text reports, with MLflow tracking and an offline evaluation harness.

   Python

2. qwen3-financial-sft-unsloth-dpo qwen3-financial-sft-unsloth-dpo Public

   Unsloth QLoRA SFT then DPO on Qwen3-4B for financial QA — MLflow-tracked, type-aware evaluation surfacing a real SFT-vs-DPO tradeoff the aggregate score hid.

   Python

3. policy-gradient-from-scratch policy-gradient-from-scratch Public

   Policy-gradient RL built up from REINFORCE to GRPO in plain PyTorch, no RL library — makes the one shared idea across the algorithm family visible, and what each successive method actually adds.

   Python

4. kafka-grid-intelligence kafka-grid-intelligence Public

   Real-time GB grid-stress prediction over Kafka, with a model-skill audit (persistence vs periodicity)

   HTML

5. nhs-care-access-agent nhs-care-access-agent Public

   Evaluation-first, model-agnostic NHS care-access agent over the nhs-intelligence-mcp tool layer

   Python

6. financial-rag-langgraph financial-rag-langgraph Public

   Python