Quantitative Researcher & Open-Source Author | Risk Magazine Quant of the Year 2024

Focused on systematic strategies, portfolio optimisation, stochastic volatility modelling, and robust statistical methods. Currently Global Head of Quantitative Analytics at LGT Private Banking. Co-originator of the Robust Optimisation of Strategic and Active Asset Allocation (ROSAA) framework and the Karasinski-Sepp log-normal beta stochastic volatility model.

For publications, speaking, and full background → artursepp.com

For contributors: OSS documentation authoring standard covers article templates, authorship, equations, references, and reproducible analytics across the packages.

For bounded starter tasks, review expectations and automation-assisted contributions, see contributing to the OSS portfolio stack.

Over 20 years of building quantitative models — across equity, credit and rates derivatives on the sell-side, a systematic CTA, market-neutral crypto/DeFi, and now multi-asset private banking — one pattern holds: volatility regimes migrate across asset classes, and models that feel robust fail at the worst moment. These open-source packages are my working answer, spanning the full quant workflow from market data — prices, fundamentals and point-in-time option chains — to signal generation, factor modelling, and portfolio construction, through to performance measurement of private assets.

Developed alongside my published research, these packages provide runnable implementations of the methods described in the papers.

New to the ecosystem? Start with qis for analytics and reporting, optimalportfolios for portfolio construction, or stochvolmodels for volatility modelling.

For three offline paths using released packages, see first workflows for the OSS portfolio stack: supplied weights to a factsheet, constrained allocation, and sparse factor estimation.

Package	Concept	Version	Stars	Forks	Total Downloads
Portfolio Construction, Factor Models, Backtest Reporting
QuantInvestStrats (qis) · Docs	Performance analytics, backtesting and factsheets
OptimalPortfolios (optimalportfolios) · Docs	Portfolio optimisation and rolling backtests
factorlasso (factorlasso) · Docs	Sparse loading estimation and factor covariance
FactorModels (factormodels) · Docs	Factor construction, point-in-time risk and GLS premia
Volatility and Option Modelling
StochVolModels (stochvolmodels) · Docs	Stochastic volatility pricing and calibration
OptionChainAnalytics (option-chain-analytics) · Docs	Point-in-time option-chain data and queries
VanillaOptionPricers (vanilla-option-pricers) · Docs	Vectorised BSM and Bachelier pricing
Dynamic Trading Strategies
TrendFollowingSystems (trendfollowing) · Docs	Closed-form trend-following analytics
GoalBasedAllocation (goal-based-allocation) · Docs	Goal-based allocation with wealth floors
Illiquid Private Assets
privateassets (privateassets) · Docs	Multi-factor PME for private assets
Data
BloombergFetch (bbg-fetch) · Docs	Bloomberg data in pandas DataFrames

Stars, forks and versions use live badges. Total downloads use Pepy dashboard counts, refreshed daily and rounded to thousands (K); counts below 1,000 are shown in full. Last successful download refresh (UTC): .

The packages support two complementary research workflows. Arrows show supplied inputs, results or analytics; they do not imply a package import dependency.

factormodels defines and constructs the financial factor model, calls factorlasso to estimate loadings and covariance, and exposes completed model outputs for downstream use. optimalportfolios consumes supplied covariance and return inputs. privateassets can use constructed factor histories and estimates fund exposures from cash flows and NAVs. trendfollowing uses QIS directly for strategy analytics. QIS also provides risk analysis, backtesting and reporting across the workflow. See the FM handoff to OP and QIS and private-asset workflow.

Vanilla pricing, point-in-time option chains and QIS analytics feed stochastic volatility valuation and calibration.

Optional market data: bbg-fetch supplies Bloomberg market data for either workflow. The stack is data-provider independent: packages accept supplied data, and bbg-fetch is optional throughout.

goal-based-allocation is a standalone research library within the broader ecosystem.

qis supplies shared analytics; factormodels constructs the financial factor model and coordinates statistical estimation through factorlasso. optimalportfolios consumes prepared risk and return inputs for portfolio construction.

qis provides Python tools for financial-data visualisation, performance reporting, and quantitative-strategy analysis. It is the analytics and reporting engine behind optimalportfolios, trendfollowing and option-chain-analytics.

pip install qis

Features:

• Backtesting engine for externally computed weights with provided instrument price, carry and cost data
• Performance reporting and factsheets: risk-adjusted tables, benchmark regressions and attribution, for multi-asset, strategy, strategy vs benchmark and multi-strategy
• Visualisation layer for financial time series built on matplotlib/seaborn

Implementation of optimisation analytics for constructing and backtesting optimal portfolios in Python. Companion code to Sepp (2023) and Sepp, Ossa & Kastenholz (2026).

pip install optimalportfolios

Features:

• Risk budgeting, alpha-focused and benchmark-constrained optimisers
• Backtesting frameworks for roll-forward analysis with handling of incomplete and missing data
• Performance attribution

Sparse factor model estimation with sign-constrained LASSO, prior-centred regularisation, hierarchical clustering group LASSO (HCGL) and factor-clustering group LASSO (FCGL), with integrated factor covariance assembly. Companion code to Sepp, Ossa & Kastenholz (2026) and Sepp, Hansen & Kastenholz (2026).

pip install factorlasso

Features:

• Sign-constrained LASSO and Group LASSO via CVXPY, with prior-centred regularisation (shrink toward β₀, not zero)
• Hierarchical Clustering Group LASSO (HCGL) and Factor-Clustering Group LASSO (FCGL) with auto-discovered groups
• NaN-aware estimation for variables with different history lengths
• Consistent factor covariance assembly (Σ_y = β Σ_x β' + D), scikit-learn compatible API (fit / predict / score)

Multiasset factor construction and point-in-time risk models built on qis and factorlasso, with GLS factor premia and capital market assumption diagnostics.

pip install factormodels

Features:

• Factor definitions and construction from tradable baskets, tracked indices and observed levels
• Completed model snapshots with labelled loadings, factor, residual and asset covariance, information cutoffs and provenance
• GLS factor premia inferred from supplied forecasts, with uncertainty and capital market assumption diagnostics
• Prepared outputs for portfolio optimisation and export to QIS risk analytics; see the completed model guide

Python implementation of pricing analytics and Monte Carlo simulations for stochastic volatility models including the Karasinski-Sepp log-normal beta SV model and the Heston model. Companion code to Sepp & Rakhmonov (2023) and Sepp & Rakhmonov (2025).

pip install stochvolmodels

Features:

• Karasinski-Sepp log-normal beta SV model, with the Heston model as benchmark
• Factor Heath-Jarrow-Morton framework for rates with log-normal stochastic volatility
• Analytical valuation of European call and put options, and Monte Carlo simulations

Python implementation of vectorised pricers and implied volatility fitters for vanilla options under Black-Scholes-Merton and Bachelier models. The pricing kernel of option-chain-analytics.

pip install vanilla-option-pricers

Features:

• Black-Scholes-Merton log-normal and Bachelier normal option pricing
• Vectorised implied volatility fitters
• Numba-accelerated implementation

Point-in-time option-chain containers, feed normalisation, chain reconstruction, queries and visualisation. The public data-container layer for empirical option research: pricing and implied-volatility inversion are delegated to vanilla-option-pricers, time-series and plotting utilities to qis.

pip install option-chain-analytics

Features:

• Timezone-aware point-in-time containers (OptionsDataDFs, SlicesChain, ExpirySlice) with exact-time chain reconstruction and no look-ahead
• Provider adapters behind optional extras — CBOE fitted chains, Deribit/Tardis crypto histories, ThetaData EOD equity/ETF reports, Bloomberg via bbg-fetch — normalised to one schema with resumable Parquet caches
• ATM-volatility and delta-skew queries, rolling vol and skew time series, and multi-page PDF chain reports

Replication package for The Science and Practice of Trend-Following Systems. Companion code to Sepp & Lucic (2026).

pip install trendfollowing

Features:

• Closed-form expected return, Sharpe ratio, skewness, and turnover of trend-following systems under white noise, AR(1), and ARFIMA processes, verified by Monte Carlo
• Three complete system implementations: European, American, and Time Series Momentum (TSMOM)
• 84-contract futures dataset spanning 1959–2026

Analytical Laplace-transform framework for dynamic mean-variance portfolio allocation under regime-switching jump-diffusions with absorbing wealth floors. Companion code to Sepp (2026).

pip install goal-based-allocation

Features:

• Riccati ODE system for MV-optimal policy with regime-dependent coefficients
• Terminal wealth density decomposition (survived + floor atom + overshoot), with exact buy-and-hold moments via matrix exponential
• Investment opportunity set construction with endogenous de-risking glide paths, validated against a Monte Carlo simulator

Multi-factor, money-weighted PME for private-asset cash flows: generalises Direct Alpha, KS-PME and GPME from a single benchmark to a tradable multi-factor deflator, with the classical measures shipped alongside for comparison on the same cash flows.

pip install privateassets

Features:

• Fund reporting to alpha in one call: NAV-implied returns → AR(1) unsmoothing with bootstrap bias correction → sign-constrained factor betas → multi-factor deflator → per-vintage and capital-weighted alpha with bootstrap intervals
• Point-in-time covariance with no look-ahead, enforced by tests; provenance (versions, seed, specification) travels with every result
• Classical single-benchmark measures (Direct Alpha, KS-PME, GPME) alongside the multi-factor versions

bbg-fetch retrieves Bloomberg prices, implied volatilities, option chains, and fundamentals into pandas DataFrames through blpapi. It is an optional data source for the provider-independent analytics stack.

pip install bbg-fetch

Features:

• Prices, implied vols, option chains and fundamentals
• Direct blpapi integration