Data & AI skills

Add Arize AX tracing to an LLM application, recording how model calls and tool executions connect. Inspect the existing code, select suitable integrations, add instrumentation, and verify that the application runs and traces reach Arize.

• Setting up tracing for an LLM application
• Recording tool inputs and outputs in an agent
• Adding Arize export to an existing OpenTelemetry setup

Choose SGLang Diffusion settings for a specific model, GPU, and memory budget. Compare generation speed, memory use, and output quality to select command-line options, including CPU offloading, multiple GPUs, caching, and reduced precision.

• Choosing generation settings for a specific model and GPU
• Reducing GPU memory use to fit a memory budget
• Comparing performance settings across multiple GPUs

Measure denoising latency in SGLang diffusion models and identify slow operations with profiler traces. Compare performance and output quality on the same GPUs, check existing optimizations, and gather evidence before starting low-level kernel work.

• Measuring diffusion denoising latency
• Ranking slow operations from profiler traces
• Checking whether existing optimizations address a bottleneck

Analyze LLM serving traces to identify GPU time usage, opportunities to run operations concurrently, and known patterns for combining operations. Produce three consistent tables from saved traces or live captures on supported servers.

• Identify GPU operations that dominate inference time
• Compare input processing and token generation
• Trace GPU operations back to Python code

Quantize diffusion transformers with NVIDIA ModelOpt, using FP8 or NVFP4 lower-precision numbers, and prepare the saved weights for SGLang Diffusion. Follow a workflow for loading, checking output quality, and comparing speed under matching conditions.

• Preparing ModelOpt FP8 weights for SGLang Diffusion
• Checking quality after diffusion model quantization
• Comparing BF16 and quantized generation speed

Analyze investment returns and portfolio risk using loss thresholds, average losses beyond those thresholds, and peak-to-trough declines. Combine simulated price paths with historical and hypothetical market shocks to produce a risk report and control recommendations.

• Examining risks behind backtest results
• Comparing portfolio losses under market shocks
• Measuring drawdowns and time to recovery

Compare Chinese A-share sectors using business conditions, price trends, valuation, and fund flows. The skill provides scoring and ranking frameworks to organize sector allocation proposals and trace how demand or cost changes affect connected industries.

• Compare business conditions across Chinese A-share sectors
• Rank sectors by price trends, earnings changes, and fund flows
• Trace cost and demand changes through industry supply chains

Browse registered trading strategies and compare evidence from backtests, which simulate trades on historical data. Filter by market conditions, trade counts, risk-adjusted returns, and cost feasibility, inspect evidence age, and rebuild the evidence cache from local runs.

• Browsing registered trading strategies
• Finding strategies with evidence for falling or rising markets
• Checking a strategy's evidence quality and age

Analyze price and volume data using trend indicators, overbought and oversold measures, and volume changes. Combine these three perspectives into a single long, short, or stand-aside signal, with adjustable calculation periods and thresholds.

• Combine several technical indicators into one signal
• Check trends alongside overbought and oversold conditions
• Adjust indicator periods and thresholds

Turn broker trade exports into a report on holding periods, trading frequency, win rate, and realized profit or loss. Diagnose patterns such as holding losers longer, overtrading, chasing rising prices, and fixation on particular price levels, with numerical evidence.

• Reviewing win rate, profit or loss, and holding periods
• Checking trading habits with numerical evidence
• Comparing results by month, symbol, or market

Generate trading signals from historical volatility, a measure of past price fluctuations. Rank its current level against recent history, signal long positions at low levels, and exit or short at high levels while retaining positions between thresholds.

• Build signals from historical volatility levels
• Adjust calculation windows and percentile thresholds
• Set annualization for cryptocurrency data

Retrieve U.S. company filings and financial series from SEC EDGAR. Resolve stock tickers to company identifiers, list recent reports with document links, and fetch reported metrics such as revenue or assets without an API key.

• Find a company's recent annual reports and document links
• List quarterly reports and material event filings
• Retrieve reported revenue or net income series

Collect financial discussions from Twitter/X, Telegram, Discord, and Reddit, then measure sentiment and discussion volume. Use the resulting data to study market attention, distinguish author groups, and test whether social signals relate to future returns.

• Compare bullish and bearish sentiment around a stock
• Detect unusual increases in discussion volume
• Track discussion in crypto project communities

Analyze crypto perpetual futures funding payments and spot–futures price differences to assess carry trades. Compare exchanges, identify crowded positions, and summarize potential returns alongside risks such as funding reversals and liquidation.

• Compare funding rates across crypto exchanges
• Estimate annualized spot–futures price differences
• Assess carry trade returns and risks

Build a pair trading strategy for two related instruments using their price ratio. Measure how far the ratio moves from its recent average, generate opposite buying and selling signals, and close positions when the ratio returns near that average.

• Test a pair trade between two stocks in the same industry
• Build a BTC/ETH price ratio strategy
• Set entry and exit thresholds for a pair trade

Apply statistical tests to financial time series, model changing volatility, and diagnose regression problems. Estimate uncertainty in strategy metrics through bootstrap resampling, which repeatedly samples existing data, and organize findings into a statistical testing report.

• Check whether price and return series are statistically stable
• Test long-term relationships between two assets
• Diagnose regression problems in factor research

Rank stocks using several indicators, such as recent returns and trading volume, combined on a common scale. Build equally weighted portfolios from the highest scores, or combine existing factors into strategies with buying and selling positions.

• Rank stocks using several investment indicators
• Build an equally weighted portfolio of top-ranked stocks
• Combine Alpha Zoo factors into a trading strategy

Provides guidance for retrieving China A-share prices and trading volumes through the TDX protocol. The source is currently retired from automatic routing because servers stopped responding, with explicit selection retained for possible future recovery.

• Reviewing the retired data source and its limitations
• Checking daily and intraday data retrieval methods
• Understanding ticker formats and output columns

Retrieve cryptocurrency prices, candlesticks, order books, funding rates, and open interest from OKX’s public API. Use spot, derivatives, and index data for market research without registering an account or configuring authentication.

• Checking cryptocurrency prices and trading volume
• Retrieving historical candlestick data
• Examining funding rates and open interest

Retrieve financial data through Tushare using Python. Find relevant APIs for Chinese stocks, ETFs, funds, bonds, economic indicators, and news, then use documented parameters to obtain pandas DataFrames, tables ready for further analysis.

• Retrieving lists of listed Chinese stocks
• Collecting ETF prices and fund size data
• Checking fund holdings and manager information

Retrieve price and trading volume data for US, Hong Kong, and Canadian stocks, ETFs, and indices through Yahoo Finance. Also gather company profiles, financial statements, dividends, and ownership data for market research and historical strategy testing.

• Collect historical prices across several stock markets
• Review company financial statements and ownership
• Check dividend and stock split histories

Generates buy, sell, or stand-aside signals from calendar patterns using price and volume data. Supports configurable months and weekdays, including combined conditions, for testing seasonal trading ideas against historical data.

• Testing trading rules based on particular months
• Adding weekday conditions to a seasonal strategy
• Combining month and weekday signals

Assess market sentiment using fear and greed measures, options activity, margin trading, foreign investment flows, and social media discussion. Combine indicators on a common scale to produce a dashboard and explain signs of excessive optimism or fear.

• Create a market sentiment dashboard
• Compare Put-Call Ratio with VIX
• Review Chinese A-share margin activity and foreign investment flows

Generate trading signals using Smart Money Concepts, a framework for interpreting institutional trading behavior. Analyze trend continuation, reversals, price gaps, and order concentration zones, then combine structural signals and gaps to determine a trading direction.

• Build signals from trend continuation and reversal patterns
• Filter structural signals using price gaps
• Adjust swing detection and closing-price break rules