How aaditya-anand11/-optiq-flow works
Project Type and Purpose
• OptiQ-Flow is a domain-specific quantum-enabled data pipeline plus an interactive dashboard. It is not a traditional “frontend/backend” web app; rather it is a scientific application with:
• Data ingestion modules (historical or synthetic market data)
• Algorithmic layers (QUBO formulation, QAOA solver with/without error mitigation, classical baseline)
• Visualization & dashboard (Streamlit, notebooks)
• Purpose: solve Markowitz portfolio optimization via QAOA on NISQ hardware, quantify error mitigation benefits, and present results interactively.
File Structure Analysis
• Top-level files:
• app.py: Streamlit entry point
• config.py: global settings (e.g., IBM token, hardware selection)
• notebooks/: exploratory Jupyter notebook
• requirements.txt / .github/: CI tests, linting
• src/ contains all core logic organized by responsibility:
• portfolio_data.py (data ingestion & preprocessing)
• qubo_formulation.py (QUBO/Ising mapping)
• qaoa_optimizer.py (circuit building & variational loop)
• error_mitigation.py (readout calibration, ZNE)
• classical_baseline.py (cvxpy Markowitz solver)
• visualization.py (Plotly/Matplotlib helpers)
• tests/: unit tests for QAOA logic
README Insights
• Tech stack spans Qiskit (quantum SDK, Aer simulator, IBM Runtime), SciPy/COBYLA, cvxpy, pandas/numpy, Plotly/Matplotlib, Streamlit.
1. Data Source
• Yahoo Finance or Synthetic Data (possibly an external API)
2. Data Processing Module
• src.portfolio_data: fetch, clean, compute returns & covariance matrices
3. QUBO/Ising Mapper
• src.qubo_formulation: portfolio_to_qubo(), qubo_to_ising()
4. Solver Layer (branch into two parallel sub-flows):
• QAOA Runner (no mitigation) – raw execution on hardware/simulator
• QAOA Runner (with mitigation) – includes readout calibration & zero-noise extrapolation
5. Classical Baseline
• src.classical_baseline: cvxpy mean-variance solution
6. Comparison & Visualization
• app.py / Streamlit UI for interactive dashboard
• notebooks/analysis.ipynb for static walkthrough
7. External Services & Config
• IBM Quantum Runtime (hardware endpoints)
• Environment (.env) for IBM token
• Data Source → Data Processing → QUBO Mapper → Solver Layer (split)
• Solver Outputs + Classical Baseline → Comparison & Visualization → Streamlit / Notebook
• config.py injects parameters into all modules
• Error Mitigation module hooks into the QAOA Runner (with mitigation branch)
c. Architectural Patterns
• Pipeline pattern: sequential data transformation steps
• Branching for A/B runs (raw vs. mitigated)
• Modular separation by concern (ETL, formulation, compute, visualization)
• Dependency injection via config.py and environment variables
d. Key Technologies & Frameworks
• Qiskit (SDK, Aer simulator, IBM Runtime)
• SciPy optimizers (COBYLA, SLSQP)
• cvxpy (classical baseline)
• pandas / NumPy (data handling)
• Plotly / Matplotlib (visualization)
• Streamlit (dashboard)
• GitHub Actions (.github/workflows/tests.yml) for CI