Portfolio Optimization

Portfolio Optimization with NVIDIA cuOpt

<!-- SPDX-FileCopyrightText: Copyright (c) 2023-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved. SPDX-License-Identifier: Apache-2.0 -->

Purpose

Build and analyze quantitative portfolios with NVIDIA-accelerated Mean-CVaR and Mean-Variance optimization. Use the portfolio_optimization package to compute returns, generate KDE scenarios for CVaR, solve variance-cap Markowitz allocations as SOCP/QCQP problems with the cuOpt GPU solver, trace an efficient frontier, backtest portfolios, and run rebalancing workflows from price data.

When to Use

Use this skill when the task is to:

• Build or optimize a Mean-CVaR portfolio from stock prices.

• Allocate weights across tickers while controlling downside CVaR risk.

• Solve Mean-Variance or Markowitz allocations with a hard variance or volatility cap using cuOpt SOCP/QCQP support.

• Plot or inspect an efficient frontier for a portfolio universe.

• Produce a weights-by-risk-aversion table.

• Backtest an optimized portfolio against benchmarks.

• Rebalance a portfolio on a schedule or drift trigger.

• Run workflows on an S&P 500, S&P 100, Dow 30, or user-supplied price dataset.

Common trigger phrases include "optimize my portfolio", "build a CVaR portfolio", "use cuOpt to optimize these tickers", "solve with cuOpt", "plot the efficient frontier", "show weights by risk aversion", "backtest this allocation", "rebalance monthly", "analyze my holdings with CVaR", "compare allocations", "reduce downside risk", "construct an allocation", "assess allocation options", "stress-test my holdings", "evaluate downside-risk exposure", "review my holdings under weight caps", "compare benchmark portfolios", "simulate CVaR scenarios", "screen portfolio risk", "optimize holdings under constraints", "solve a variance-cap portfolio", "use SOCP", "set a volatility cap", and "find a lower-risk allocation".

Do not use it for generic finance summaries, price forecasting, neural-network training, vehicle routing, or non-portfolio optimization.

Prerequisites

• Python environment with the installed portfolio_optimization package.

• NVIDIA GPU runtime with cuOpt and cuML installed. Mean-Variance SOCP workflows require a cuOpt build with QCQP/SOCP support, such as the 26.06 line or newer.

• CUDA extra matching the host and workflow: uv sync --extra cuda12 for full cuOpt/cuML 26.06 on CUDA 12, uv sync --extra cuda13 for the current full CUDA 13 stack, or uv sync --extra cuda13-socp for CUDA 13 SOCP-only validation with cuOpt 26.06.

• cvxpy exposing cp.CUOPT.

• Network access on first run if the default price CSV must be downloaded.

Setup

This skill drives the installed portfolio_optimization package. A ready environment can come from the Brev launchable or from the NVIDIA-AI-Blueprints/portfolio-optimization repository after installing the matching CUDA extra.

In packaged agent/eval sandboxes, portfolio_optimization may be available through PYTHONPATH rather than as a separately published wheel. Verify the local package with python -c "import portfolio_optimization" before declaring it missing. Do not pip install portfolio_optimization; do not reimplement the example workflows from scratch, and do not replace the package APIs with generic pandas/scipy/cvxpy portfolio code.

For concrete implementation details, use references/workflows/agent_recipes.md as the source of truth. It contains exact working shapes for loading prices, preparing returns, solving with cuOpt, building a 25-point frontier, backtesting against equal weight, and calling the rebalancer.

The default dataset is data/stock_data/sp500.csv. It is gitignored. Before a first-run download, tell the user this fetches public market data through the package's yfinance data helper and ask them to confirm:

import cvxpy as cp
from portfolio_optimization.cvar_parameters import CvarParameters
from portfolio_optimization.utils import download_data

download_data("data/stock_data", datasets=["sp500"])
CVAR_SOLVER_SETTINGS = {"solver": cp.CUOPT, "verbose": False, "solver_method": "PDLP"}
cvar_params = CvarParameters(
    w_min=0.0, w_max=1.0,
    c_min=0.0, c_max=0.0,
    risk_aversion=1.0, confidence=0.95,
)

Instructions

Briefly state the defaults being applied before execution, then use these guardrails:

1. Load data/stock_data/sp500.csv; if it is missing, ask before downloading sp500 with portfolio_optimization.utils.download_data. Do not glob, substitute, or fabricate price data.

2. Validate user CSVs before solving: require a date-like index or first date column, numeric ticker columns, at least 60 rows after date filtering, and at least one requested ticker. If the user gives start/end dates, slice the price DataFrame before returns computation and report the retained date range. Filter tickers on the price DataFrame before returns are computed. regime_dict does not take a ticker field.

3. Compute LOG returns with utils.calculate_returns(...).

4. For Mean-CVaR tasks, generate scenarios with cvar_utils.generate_cvar_data(...), KDE, and KDESettings(device="GPU"). For Mean-Variance SOCP variance-cap tasks, do not generate CVaR scenarios; use the returns_dict directly after LOG return computation.

5. For ordinary Mean-CVaR portfolio requests, define CvarParameters with explicit w_min and w_max, and set c_min=0.0 and c_max=0.0 so the result is fully invested instead of 100 percent cash.

6. For variance-cap, volatility-cap, Markowitz, SOCP, or QCQP requests, define MeanVarianceParameters with var_limit set to a positive variance bound, c_min=0.0, c_max=0.0, and L_tar=1.0 for long-only fully invested allocations. If the user gives a volatility cap, square it before assigning var_limit.

7. Build cvar_optimizer.CVaR(returns_dict, cvar_params) for Mean-CVaR tasks. Build mean_variance_optimizer.MeanVariance(returns_dict, mean_variance_params, api_settings=ApiSettings(api="cuopt_python")) for direct cuOpt Mean-Variance SOCP tasks.

8. Solve with NVIDIA cuOpt only. For CVaR, verify hasattr(cp, "CUOPT") and str(cp.CUOPT) in {str(s) for s in cp.installed_solvers()}, then pass CVAR_SOLVER_SETTINGS to every single-shot solve or looped frontier solve. For direct Mean-Variance SOCP, verify the cuopt Python package is importable and call the optimizer with api="cuopt_python"; cuOpt auto-selects the barrier method for quadratic constraints. Never fall back to CLARABEL, SCS, ECOS, or another CPU solver. If cuOpt is absent, finish validation/setup and report that the GPU/cuOpt runtime is missing instead of fabricating a CPU result.

9. For custom constraints, map user requests to the appropriate parameter model: CVaR risk controls to CvarParameters, variance or volatility caps to MeanVarianceParameters.var_limit, weight caps to w_min/w_max, risk appetite to risk_aversion, confidence level to confidence, and cash allowance to c_max. Treat cardinality plus SOCP as unsupported unless the package exposes explicit mixed-integer conic support.

10. If the user omits a benchmark for backtesting, use an equal-weight portfolio over the same tickers. If the user omits a constraint, keep the defaults table values and briefly restate consequential assumptions before solving.

11. Deliver weights sorted by allocation, cash weight, expected return, solver label (cuOpt GPU), and the risk metric used: CVaR for Mean-CVaR or realized variance plus var_limit for SOCP. Include any requested frontier figure, weights table, backtest metrics, or rebalancing schedule. For tables, include tickers as columns or rows with decimal weights and percentages; for plots, preserve the figure returned by the package instead of redrawing from scratch.

12. For report-grade answers, include evidence that the requested workflow actually ran. For an efficient frontier, state len(results_df) and use the requested ra_num (25 unless the user specifies otherwise). For a variance-cap SOCP solve, report result_row["solver"], realized variance, the requested var_limit, and confirm realized variance is at or below the cap. For a weights table, expand results_df["weights"] into ticker columns and include cash plus risk_aversion. For a backtest, include mean portfolio return, sharpe, sortino, and max drawdown for both optimized and benchmark portfolios. For rebalancing, include results_dataframe, re_optimize_dates, and the tail of cumulative_portfolio_value.

Canonical Workflow Skeleton

Start applicable portfolio optimization tasks from this shape and adapt only the requested output. For complete copyable functions, read references/workflows/agent_recipes.md before writing custom code.

Mean-CVaR workflow

import cvxpy as cp
import pandas as pd

from portfolio_optimization import backtest, cvar_optimizer, cvar_utils, rebalance, utils
from portfolio_optimization.cvar_parameters import CvarParameters
from portfolio_optimization.portfolio import Portfolio
from portfolio_optimization.settings import KDESettings, ReturnsComputeSettings, ScenarioGenerationSettings

if not hasattr(cp, "CUOPT") or str(cp.CUOPT) not in {str(s) for s in cp.installed_solvers()}:
    raise RuntimeError("cuOpt GPU solver is required; do not substitute a CPU solver.")

CVAR_SOLVER_SETTINGS = {"solver": cp.CUOPT, "verbose": False, "solver_method": "PDLP"}

prices = utils.get_input_data("data/stock_data/sp500.csv")
returns_dict = utils.calculate_returns(
    prices,
    regime_dict=None,
    returns_compute_settings=ReturnsComputeSettings(return_type="LOG"),
)
returns_dict = cvar_utils.generate_cvar_data(
    returns_dict,
    ScenarioGenerationSettings(
        fit_type="kde",
        kde_settings=KDESettings(device="GPU"),
    ),
)
cvar_params = CvarParameters(
    w_min=0.0,
    w_max=1.0,
    c_min=0.0,
    c_max=0.0,
    risk_aversion=1.0,
    confidence=0.95,
)
optimizer = cvar_optimizer.CVaR(returns_dict, cvar_params)
result, optimal_portfolio = optimizer.solve_optimization_problem(
    solver_settings=CVAR_SOLVER_SETTINGS,
    print_results=False,
)

Mean-Variance SOCP workflow

import importlib.util
import numpy as np

from portfolio_optimization import mean_variance_optimizer, utils
from portfolio_optimization.mean_variance_parameters import MeanVarianceParameters
from portfolio_optimization.settings import ApiSettings, ReturnsComputeSettings

if importlib.util.find_spec("cuopt") is None:
    raise RuntimeError("cuOpt Python API is required; do not substitute a CPU solver.")

prices = utils.get_input_data("data/stock_data/sp500.csv")
returns_dict = utils.calculate_returns(
    prices,
    regime_dict=None,
    returns_compute_settings=ReturnsComputeSettings(return_type="LOG"),
)
weights = np.ones(len(returns_dict["tickers"])) / len(returns_dict["tickers"])
var_limit = float(weights @ returns_dict["covariance"] @ weights) * 1.05
mean_variance_params = MeanVarianceParameters(
    w_min=0.0,
    w_max=1.0,
    c_min=0.0,
    c_max=0.0,
    L_tar=1.0,
    var_limit=var_limit,
)
optimizer = mean_variance_optimizer.MeanVariance(
    returns_dict,
    mean_variance_params,
    api_settings=ApiSettings(api="cuopt_python"),
)
result, optimal_portfolio = optimizer.solve_optimization_problem(print_results=False)
realized_variance = float(
    optimal_portfolio.weights @ returns_dict["covariance"] @ optimal_portfolio.weights
)

For an efficient frontier or weights table, call:

results_df, fig, ax = cvar_utils.create_efficient_frontier(
    returns_dict,
    cvar_params,
    CVAR_SOLVER_SETTINGS,
    ra_num=25,
    show_plot=False,
    show_discretized_portfolios=False,
    benchmark_portfolios=False,
    print_portfolio_results=False,
)
weights_table = pd.DataFrame(results_df["weights"].tolist(), index=results_df.index)

For a benchmark backtest, wrap the solved allocation in Portfolio(name="cuOpt Optimal", tickers=returns_dict["tickers"], weights=optimal_portfolio.weights, cash=optimal_portfolio.cash), create an equal-weight Portfolio over the same returns_dict["tickers"], then use backtest.portfolio_backtester(..., test_method="historical").backtest_against_benchmarks(...). The backtester returns (backtest_results, ax).

For monthly rebalancing, write the price DataFrame to a CSV path first. Instantiate rebalance.rebalance_portfolio(dataset_directory=<csv_path>, ...) with re_optimize_criteria={"type": "drift_from_optimal", "threshold": 0, "norm": 1} and call re_optimize(transaction_cost_factor=..., plot_title="Monthly Rebalancing"). The rebalancer returns (results_dataframe, re_optimize_dates, cumulative_portfolio_value).

Data and Defaults

The default S&P 500 file is a historical snapshot and can omit current constituents. User-supplied CSVs should be date-indexed price tables with ticker columns, compatible with utils.get_input_data. If requested tickers are absent, drop them, report the omissions, and continue with available columns unless the user explicitly asks you to fetch other data.

Key APIs

Use the package APIs instead of reimplementing portfolio math or simulation loops. portfolio_optimization helpers return flat objects: returns_dict has keys such as returns, mean, covariance, and tickers; do not index it as returns_dict["regime_1"]. solve_optimization_problem(...) returns (result_row, portfolio), not a nested result dictionary.

• Returns: utils.calculate_returns(input_dataset, regime_dict, returns_compute_settings).

• Regime filter: regime_dict is None or {"name": "...", "range": ("YYYY-MM-DD", "YYYY-MM-DD")}; it is not keyed by regime name and does not contain tickers.

• Scenarios: cvar_utils.generate_cvar_data(returns_dict, scenario_generation_settings) for Mean-CVaR only.

• CVaR optimizer: cvar_optimizer.CVaR(returns_dict, cvar_params).

• Mean-Variance SOCP optimizer: mean_variance_optimizer.MeanVariance(returns_dict, mean_variance_params, api_settings=ApiSettings(api="cuopt_python")).

• CVaR solve: result_row, portfolio = cvar_problem.solve_optimization_problem(solver_settings=CVAR_SOLVER_SETTINGS, print_results=False).

• SOCP solve: result_row, portfolio = mean_variance_problem.solve_optimization_problem(print_results=False).

• Efficient frontier: cvar_utils.create_efficient_frontier(returns_dict, cvar_params, solver_settings=CVAR_SOLVER_SETTINGS, ra_num=25). The returned results_df includes metrics, a weights dict column, and cash.

• Portfolio: Portfolio(name="", tickers=None, weights=None, cash=0.0, time_range=None); pass tickers and a flat array-like weights aligned to those tickers.

• Backtest: create portfolio.Portfolio objects for the optimized allocation and each benchmark; for an equal-weight benchmark, use weights of 1 / len(tickers) and cash=0.0, then call backtest.portfolio_backtester(test_portfolio, returns_dict, risk_free_rate=0.0, test_method="historical", benchmark_portfolios=[...]).backtest_against_benchmarks(...).

• Rebalance: rebalance.rebalance_portfolio(...) requires dataset_directory to be a CSV path, not a DataFrame. Call re_optimize(...); it returns (results_dataframe, re_optimize_dates, cumulative_portfolio_value).

• Settings models: ReturnsComputeSettings, ScenarioGenerationSettings, KDESettings, ApiSettings, CvarParameters, and MeanVarianceParameters.

Examples

• "Build the optimal portfolio from the S&P 500": load prices, compute LOG returns, generate GPU KDE scenarios, set long-only fully invested CvarParameters, solve with cuOpt, and report diversified weights plus return/CVaR.

• "Solve a variance-cap portfolio with SOCP": load prices, compute LOG returns, set MeanVarianceParameters(var_limit=...), solve with direct api="cuopt_python", and report expected return, realized variance, var_limit, and weights.

• "Plot the efficient frontier": call create_efficient_frontier(...), return results_df, and show or save the figure as requested.

• "Give me weights by risk aversion": expand results_df["weights"] into a per-asset table.

• "Backtest against equal weight": build the optimized and equal-weight Portfolio objects, then use the package backtester and report Sharpe, Sortino, and max drawdown.

• "Backtest monthly rebalancing": configure rebalance_portfolio with the drift trigger above and run re_optimize(transaction_cost_factor=...).

Limitations

• Requires an NVIDIA GPU with cuOpt and cuML; CPU solvers are intentionally disallowed.

• Mean-Variance SOCP variance caps require cuOpt QCQP/SOCP support. Use the 26.06 line or newer when installing CUDA extras.

• cuda13-socp intentionally installs cuOpt without cuML because cuml-cu13 26.06 is not published yet; use it for direct SOCP/QCQP validation, not GPU KDE CVaR workflows.

• Cardinality plus SOCP is treated as unsupported unless the package exposes explicit mixed-integer conic support.

• CPU-only eval containers can still validate routing, data handling, and reporting behavior, but they cannot produce a valid cuOpt solve. In that case, report the missing GPU/cuOpt runtime explicitly.

• Default price data is a historical snapshot and may omit current constituents.

• First-run dataset download depends on network access unless the user supplies a CSV.

Troubleshooting

• Missing default CSV or FileNotFoundError: explain that the package will fetch public market data with download_data("data/stock_data", datasets=["sp500"]); run it only after user confirmation.

• SolverError or missing cp.CUOPT: install the CUDA extra matching the host and verify with python -c "import cvxpy as cp; print(hasattr(cp, 'CUOPT'), cp.installed_solvers())".

• ImportError for cuml or GPU KDE failures: confirm cuML is present with python -c "import cuml" and keep KDESettings(device="GPU"). If using cuda13-socp, this is expected for CVaR/KDE; switch to cuda12 or cuda13 for cuML workflows.

• SOCP setup fails before solving: verify the installed cuopt package is on the 26.06 line or newer and that MeanVarianceParameters.var_limit is positive.

• Ordinary optimization returns all cash: set c_max=0.0 in CvarParameters.

• Solver reports infeasible or no solution: check for contradictory bounds, too few tickers for the requested caps/cardinality, or a date filter that leaves too little data; report the smallest constraint change that would make the request feasible.

• Requested tickers are absent from the default CSV: report them and proceed with the remaining requested tickers.

• User CSV fails validation: ask for a date-indexed price table or a CSV whose first column is dates and remaining columns are numeric ticker prices; mention the minimum 60-row post-filter requirement.