Data & Analytics

基金分析与筛选：晨星评级/夏普比率/信息比率、Sharpe风格箱分析、风格漂移检测、基金经理评价、FOF组合构建、ETF选择

Factor research framework with IC/IR analysis, quantile backtesting, and factor combination. Suitable for cross-sectional factor evaluation across multiple instruments.

ETF分析：产品筛选、费率对比、跟踪误差、流动性评估、策略应用与中国市场ETF量化配置框架。

Elliott Wave Theory signal engine. Detects swing points through Zigzag, matches 5-wave impulse and 3-wave corrective structures, validates them with Fibonacci wave relationships, and generates trend-top / correction-complete signals. Pure in-house pandas implementation.

盈利预测与一致预期分析（自上而下/自下而上预测法/SUE/PEAD/分析师预期修正），捕捉业绩超预期交易机会。

Correlation and cointegration analysis — co-movement discovery, deep return-correlation analysis, sector clustering, realized correlation, Engle-Granger / Johansen cointegration, half-life, Kalman dynamic hedge ratio, cross-market linkage analysis, and pair-trading signal generation

公司事件驱动分析：并购套利价差计算、大股东增减持信号、股权激励解读、定增配股影响评估、A股ST/退市预警

Commodity analysis (oil supply-demand balance / gold pricing / copper as an economic predictor / inventory cycles / futures premium-discount structure / seasonality), generating directional commodity signals.

Candlestick pattern recognition engine, pure pandas vectorized implementation of 15 classic candlestick patterns (5 single-candle + 5 double-candle + 4 triple-candle + 1 trend confirmation), generating a composite signal from bullish/bearish pattern scores.

Browse and bench the bundled alpha zoos — prebuilt cross-sectional factor libraries (Kakushadze 101, GTJA 191, Qlib 158, Fama-French / Carhart). Use when the user asks "which alphas exist", wants metadata on a named alpha, or wants to run IC/IR on a whole zoo over a universe.

CLI harness for NSLogger — parse, filter, export, and monitor NSLogger log files (.rawnsloggerdata / .nsloggerdata)

Command-line interface for CloudAnalyzer — Agent-friendly harness for CloudAnalyzer, a QA platform for mapping, localization, and perception outputs. Supports 27 commands across 8 groups: point cloud evaluation, trajectory evaluation, ground segmentation QA, config-driven quality gates, baseline evolution, processing, visualization, and interactive REPL.

Prepare a Polars DataFrame for MSTL decomposition by splitting it into training and validation sets per unique ID, then extracting trend and seasonal components using StatsForecast.

A Python function to load and preprocess paired screenshot and HTML files from separate directories, matching them by base filename (e.g., screen_13.png with html_13.html), resizing images, and normalizing pixel values for model training.

Optimizes memory consumption during PyTorch model training by implementing mixed precision training, gradient accumulation, and efficient data loading strategies to fit within hardware constraints.

Analyzes literary texts to identify themes, devices, and POV, while constructing structured essay paragraphs with specific constraints (length, quote support, and technique analysis).

Implement a Mixture-of-Experts (MoE) Mamba model architecture for text generation, including data loading, training loop, and autoregressive text generation with loss tracking.

Scrapes TV show data (title, genres, episodes, rating) from a Next.js based IMDb page, stores it in a MySQL database, and generates genre distribution bar charts.

Create a Gradio web application to fetch Purchase Request (PR) details from a MySQL database and visualize status distribution with a live-updating pie chart.

Implements a Genetic Algorithm (GA) using DEAP to select optimal features for a classification model (e.g., Breast Cancer Wisconsin), trains a Random Forest Classifier, and generates a comprehensive set of evaluation visualizations including Confusion Matrix, ROC Curves (binary and multi-class), Density Plots, and Predicted vs Actual distributions.

Generates a comprehensive Comp/Con database entry for a mech in the Lancer TTRPG system, including detailed statistics, loadouts, core systems, traits, and creative descriptions based on the manufacturer and mech name.

Fine-tunes a pre-trained GPT-2 model on JSONL datasets (e.g., Q&A pairs) using Hugging Face Transformers. Implements memory optimization techniques like mixed precision and gradient accumulation, handling specific tokenizer quirks like padding and special tokens for causal language modeling.

Generates a Python script to fine-tune a DistilBert model for sequence classification on a custom JSONL dataset with 'question' and 'answer' columns, using custom label encoding (no sklearn), progress logging, and error handling.

Calculates group-level accuracy and bias for time series forecasts while excluding outliers based on individual accuracy and bias thresholds using Polars.