An experimental framework for optimizing equity SMAs (Separately Managed Accounts, aka standalone portfolios) for retail distribution.
Goal: optimize representative clusters of SMAs subject to account minimums in the shortest time possible.
-
Build optimization model portfolio
- modularize model components (variables, objectives and constraints)
- build model-to-use-case matrix
- build JIT model-assembler
-
Build test harness
- visualize execution waterfall
-
Test and tune Performance-Train components (and replace where needed)
- CPU
- RAM
- OS
- env
- UI
- scripts
- data
- model
- solvers
| Initial | Next up | |
|---|---|---|
| CPU | M2 | Intel Xeon E-2488 |
| RAM | 8GB | 128GB DDR 4 ECC |
| OS | macOS Sonoma 14.5 | Linux Ubuntu Server |
| Env | Docker Compose v2.27.0-desktop.2 | -- |
| Script (glue) langs | Python | Dask |
| Database | PostgreSQL + TimescaleDB | kdb+ |
| Dataframes | Pandas | Polars |
| Model langs | AMPL | Pyomo, gurobipy |
| Solvers | Gurobi (optionally + HiGHS for small problems) | -- |
| UI | Grafana |
| DB Admin | CloudBeaver |
- Yahoo Finance API
- S&P 500
- Russell 2000