Models I have trained from scratch, fine-tuned and published on Hugging Face, open for anyone to download and run.
Trisham97/llm-complexity-router
Classifies a prompt as easy, medium or hard and recommends a Claude tier (Haiku, Sonnet or Opus), packaged as a Claude Code plugin with a prompt hook, an MCP tool and per-tier worker subagents.
Logistic-regression and MLP heads trained on 61K prompts and run locally in pure NumPy (no PyTorch, no pickles). The MLP reaches 72.7% accuracy and 0.707 macro F1 on a 7.7K-prompt held-out split.
View model card on Hugging Face →QuantBridge/energy-news-classifier-ner-multitask
One DistilBERT encoder with two heads: one forward pass returns BIO named entities and multi-label news topics for energy and financial headlines.
The encoder and NER head are transferred from the custom NER model; the 10-label topic head is trained on AG News, Reuters-21578 and the Kaggle News Category corpus with BCE loss.
View model card on Hugging Face →QuantBridge/distilbert-energy-intelligence-multitask-v2
Broad-coverage NER for financial and energy intelligence with 59 entity types, from central banks and trading hubs to sanctions and regulators.
Built for structured extraction from news feeds, earnings transcripts and regulatory filings: knowledge-graph population, trading signals and geopolitical risk monitoring.
View model card on Hugging Face →QuantBridge/energy-intelligence-multitask-custom-ner
DistilBERT fine-tuned to tag nine energy-market entity types: companies, commodities, infrastructure, markets, events, countries and more.
Trained on 11K+ BIO-annotated energy and geopolitics articles. It is the backbone the multitask model reuses.
View model card on Hugging Face →QuantBridge/energy-news-classifier
Multi-label classifier that tags news across ten themes (energy, politics, trade, shipping, macro, risk and more) to turn headlines into signals.
The first model in the QuantBridge energy-intelligence series, later folded into the multitask model's classification head.
View model card on Hugging Face →