Section A
The nightly pipeline
Most quantitative funds employ research teams that test ideas as they arise. We built something different: an autonomous pipeline that tests everything, systematically, every single night. Six stages, fully automated, from raw scan to deployment decision.
1
Scan
173 archetype configurations tested against the instrument universe
Every archetype -- momentum, mean reversion, carry, volatility regime, and others -- is applied across thousands of instruments. The scan is exhaustive by design.
2
Verify
Winners go through walk-forward holdout test
Any configuration that passes the initial screen is tested on data it has never seen. This is the same holdout protocol described in our walk-forward methodology. No exceptions.
3
Correlate
Check independence from existing engines
A strategy that passes holdout but is correlated to something already in the portfolio adds risk without adding return. The correlation gate eliminates redundancy.
4
Simulate
Portfolio impact: does adding this improve Sharpe?
The candidate is added to a simulated version of the current portfolio. If the portfolio Sharpe ratio does not improve, the candidate is killed regardless of its standalone performance.
5
Deploy
Automatically added to paper trading
Survivors enter paper trading for live verification. No strategy touches real capital until it has been validated in the live market environment with zero risk.
6
Report
Results posted, cockpit updated
Every run produces a full report: what was tested, what passed, what failed, and why. The portfolio cockpit is updated automatically. By market open, the results are available.
Section B
Scale of operation
The factory is not a screening tool that filters a small list. It is a research machine that covers the full investable universe across multiple asset classes, every night, without human intervention.
Archetype Configs
173
Momentum, reversion, carry, vol regime, and more
Equities
4,784
Dynamic universe, rebalanced continuously
Crypto Perpetuals
212
Cryptocurrency perpetual futures
CME Futures
39
Contracts across commodities, rates, and indices
Scheduled Run
2:00
AM daily via launchd automation
Results Available
Open
By market open, every morning
Section C
What the factory has found
From the factory pipeline, two legs are currently deployed in the portfolio. Each was discovered autonomously, validated through holdout testing, confirmed as uncorrelated to existing engines, and deployed to paper trading without human intervention.
Factory Discovery
Vector
Vector -- discovered via the time-series momentum archetype applied to the technology sector.
Sharpe Ratio
0.80
Archetype
ts_momentum
Status
Deployed
Factory Discovery
Crosswind
Crosswind -- discovered via the cross-sectional momentum archetype across commodity instruments.
Sharpe Ratio
0.64
Archetype
xs_momentum
Status
Deployed
And the factory continues scanning. Here is the most recent run -- a snapshot of the funnel in action.
Latest Run — Selection Funnel
Zero deployed from the latest run. Correlation too high with existing legs or insufficient portfolio improvement. The system does not lower its standards.
The factory's job isn't to find strategies. It's to prove that no stone is unturned.
Section D
Archetype library
Each archetype encodes a different market hypothesis. The factory applies every archetype to every eligible instrument, generating thousands of unique configurations per run. Here are five examples from the library.
Time-Series Momentum
"Does the asset's own trend predict its future?"
Cross-Sectional Momentum
"Which assets are outperforming their peers?"
Spring
"Has the price stretched too far from equilibrium?"
Volatility Regime
"Is the current vol environment favorable?"
Carry
"What premium exists for holding this position?"
Section E
Why autonomous matters
A human researcher brings intuition and creativity. But a human researcher also brings selection bias, emotional attachment, and a limited number of hours in a day. The factory eliminates three of those constraints while preserving the fourth through its archetype design.
No selection bias
The machine tests everything, not just what a human thinks will work. Every archetype, every instrument, every configuration -- systematically.
No emotional attachment
If it doesn't pass, it gets killed automatically. No researcher lobbying for "one more test." No sunk cost fallacy. The threshold is the threshold.
Continuous
New data, new instruments, new tests every night. The market evolves. The factory evolves with it. An edge that decays is detected and removed.
Scalable
Adding a new asset class or archetype is one configuration change. The pipeline handles the rest -- scan, verify, correlate, simulate, deploy, report.
Most funds have a research team that tests ideas. We have a research machine that tests everything.
The factory is not a replacement for human judgment. It is the infrastructure that ensures human judgment is applied to the right candidates. By the time a strategy reaches a portfolio manager's attention, it has already survived six automated gates that most strategies cannot clear.