Specter

A collection of smaller fine-tuned models using Hugging Face Datasets to practice fine-tuning principles using Unsloth

No project description available.

Production-ready Shariah-compliant quant data engine. Enforces strict financial screening, calculates purification ratios, and serves clean, point-in-time signals for backtesting and strategy execution.

Build a lean, high-margin AI engineering lab that solves the $400B "Big Token" Problem for B2B SaaS companies, SMMA agencies, and Ops/Logistics teams.

LeaseGuard is an open-source, domain-adapted 8B LLM fine-tuned for commercial real estate lease extraction, multi-jurisdictional risk scoring, and automated clause redlining. Engineered for precision structured JSON output and low-latency micro-inference—delivering contract analysis at a fraction of frontier model API costs

Quant project exploring volatility-driven market regime detection using Hidden Markov Models (HMM) and VIX signals. Simulates asymmetric downside avoidance ("fear") by dynamically rotating SPY exposure into cash during panic regimes to minimize Max Drawdown and optimize Calmar Ratio