Quant & AI
Quant & AI
Quantitative finance, machine learning and model risk: research, analysis and open-source code on how models behave when markets change.
Latest analysis
-
A Rally Without Its General: Why Wall Street Rose While Nvidia Fell Nearly 13%
In five sessions, the most important stock in the AI trade lost its founder-CEO and about $580 billion of market value, and the S&P 500 still finished the week higher. What that says about how this market is wired, and…
-
Five Percent Changes Everything: Sticky Inflation, a 5% Treasury Yield and the Real Cost of the AI Build-Out
Stocks recovered part of last week’s losses. The bond market went the other way, and the bond market is the one that sets the price of every long-dated promise on Wall Street, AI included.
-
Nvidia After Huang: Four Risks the Market Is Repricing
One company’s leadership crisis moved the whole market. A look at concentration, key-person, correlation and rate risk, with the numbers behind each.
-
CoreWeave revenue jumps 168% as AI demand grows, but losses widen
CoreWeave’s third-quarter revenue rose to $3.66 billion and its backlog to $132.5 billion; the net loss widened to $709 million as capital spending surged.
-
Anthropic sets IPO price range of $70 to $78, seeking up to $64 billion
Anthropic plans to sell 825 million shares at $70 to $78, raising up to about $64.4 billion, with trading on Nasdaq expected from November 19.
-
The Election Was the Match, Not the Fuel: What Thursday’s Chip Crash Really Revealed About the AI Trade
A 12% one-day collapse in semiconductors needs more than a headline to explain it. Read the tape closely and the selloff looks less like a verdict on AI demand and more like a crowded trade meeting a memory cycle that…
Research
Model validation
Backtesting machine-learning and conventional VaR models, and whether aggregate tests hide weaknesses in stressed markets.
Tail and systemic risk
Treating extreme losses as episodes: how long they persist, how severe they become and how they spread across markets.
Machine-learning risk models
Liquidity- and regime-aware gradient-boosting Value-at-Risk for US equities.
Open-source code
ml var validation ↗
Replication code and data for When Backtests Lie, the governance-grade validation framework for machine-learning VaR.
conditional tail episode risk ↗
Research code, experiments and reproducibility materials for Conditional Tail Episode Risk (CTER).
VaR Backtesting SR11 7 ↗
Walk-forward backtest of Historical Simulation, Parametric and Student-t VaR on S&P 500 returns with Kupiec, Christoffersen and Basel traffic-light tests.
GARCH Volatility Forecasting ↗
GARCH, EGARCH and HAR-RV volatility forecasting with rolling estimation, regime analysis and forecast evaluation.
Statistical Arbitrage Pairs Trading ↗
Cointegration testing, Z-score signals, Kalman filter, Kelly sizing and a full backtest engine for pairs trading.
Order Book Microstructure Market Impact ↗
Bid-ask spread decomposition, Kyle lambda market impact, Amihud illiquidity and VWAP/TWAP execution.
Research and code are by Niraj Neupane. Working papers have not completed journal peer review unless stated. Nothing in this section is investment advice.
