Cross-Sectional Momentum
EXPERIMENTALRanks stocks by 12-month return skipping the latest month.
Caveat: Momentum suffers sharp crashes in market rebounds (−76% in the 1930s in our century-long industry test).
Methodology & evidence →AI-native quantitative research
Intelligence beyond
the market.Engineered for alpha.
A research-first platform uniting systematic strategies, AI-assisted analysis, rigorous simulation, and disciplined risk controls.
RESEARCH & PAPER TRADING ENVIRONMENT
02The intelligence network
From raw data to a reconciled paper fill — every hop explicit, typed and audited.
Select a stage to see what it does and where it lives in the code. Two stages are hard gates.
Stage 1 of 10
Market data
Daily US equity/ETF bars (Alpaca SIP), SEC filings by filing time, Form 4 insider trades, FINRA short interest, Treasury curve, Benzinga news, NY Fed reference rates. Every record carries the time it became public.
source: services/market-data
03Multi-agent research council
Seven specialists. One synthesis. Zero order authority.
Agents run on a local model, cite their sources with as-of times, and disagree openly. Deterministic models — not LLM confidence — drive ranking, sizing and risk.
CIO synthesis
ImplementedWeighs disagreements, states assumptions and confidence, produces a thesis with evidence.
Representative outputDEMO · SIMULATED
Agents agree on quality, disagree on valuation. Confidence: moderate. Thesis: HOLD.
Illustrative only — not a live recommendation for any security. Real theses cite source documents with as-of timestamps.
04Strategy universe
Ten research programs. Honest status on every one.
Lifecycle states come live from the strategy registry. Our own out-of-sample tests find no strategy that significantly beats simply holding the universe equally weighted — we publish that rather than hide it.
05Market intelligence
Every chart says where its data came from — and how old it is.
SPY SPDR S&P 500 ETF · daily
Alpaca SIP daily bars, split/dividend-adjusted · last completed session 2026-11-03 (America/New_York) · delayed, not real-time
20-day ±2σ band200-day trend (regime)
Regime read (deterministic): close is above its 200-day average. Context only — not a trading signal.
06Research to paper execution
Risk isn't a filter on the way out. It's a wall.
- 01Market data
- 02Point-in-time features
- 03Alpha models
- 04AI research council
- 05Strategy pods
- 06Portfolio construction
- 07 · HARD GATEDeterministic risk
- 08 · HARD GATEHuman authorization
- 09Alpaca paper OMS
- 10Reconciliation & ledger
Orders are proposed, risk-checked, approved by a named person (or a human-granted envelope with hard caps), re-checked at the live price, then sent to Alpaca's paper endpoint. Unknown broker outcomes are reconciled by client order ID — never blindly resubmitted.
07Risk architecture
A lattice of constraints, checked on every order.
Controls seek to manage risk; they cannot eliminate losses. The engine is deterministic, independent of any language model, and fails closed.
Current documented policyread from the risk engine
- Position
- ≤ 5% of equity per name
- Sector
- ≤ 25% per sector
- Gross / net
- ≤ 1.0× (no leverage)
- Beta
- ≤ 1.3
- Liquidity
- order ≤ 1% of 20-day volume; min $5M daily value
- Correlation
- ≤ 15% in names correlated above 0.85
- Daily loss
- 3% halts new risk
- Drawdown
- 15% engages the kill level
- Kill switch
- Fail-closed; a model can engage it, only a human releases it
- Short selling, margin, derivatives
- Disabled
08Backtesting & scientific validation
Built to find out it's wrong — before money finds out.
Walk-forward evaluationSYNTHETIC DEMONSTRATION
Synthetic curves (range 97–122) illustrate the method only — not results.
Point-in-time data
Filings count only after they were public; tests rewrite the future and prove features don't move.
Purge, embargo, locked holdout
No training label overlaps a test period; the final holdout can be opened once, by a named person.
Costs and stress
Every run is repeated at 2× costs; turnover is charged.
Uncertainty, not point estimates
Bootstrap confidence intervals, deflated Sharpe and multiple-testing correction — failed trials counted.
Survivorship disclosed
Today's large-cap universe flatters history; every result is flagged.
09The research terminal
Dense where it matters. Calm everywhere else.
DEMO DATA · PAPER
Paper NAV (demo)
$100,000.00
Risk utilization (demo)
Largest position62%
Largest sector41%
Drawdown12%
Approval queue (demo)
BUYXYZ12Risk ✓ · awaiting human
SELLABC4Risk ✓ · approved
BUYQRS40Risk ✕ · position > 5%
Ask the assistant (demo)
Does the momentum model beat equal weight?
Not significantly: 90% CI of the Sharpe difference spans zero.
Altunyan AI Hedge Fund
The next generation of investment intelligence.
Research and paper trading only. Access is invitation/approval based; registering never grants trading permission. Simulated and backtested results are hypothetical. Disclosures.
Cross-Sectional Momentum: EXPERIMENTALResidual Momentum: EXPERIMENTALQuality & Value: EXPERIMENTALTrend Following: EXPERIMENTAL