Live on Binance

AI agents that trade — backed by an ensemble of 5 ML models.

DeepAlpha invokes autonomous trading agents on any schedule — every 15 minutes, a cron expression, a specific UTC time, or the instant price breaks a level. Agents reason with a production ML stack and execute against your broker.

Request accessSee how it works →
1.6K
Agent runs
5
Models in ensemble
21
Crypto assets
200+
Indicators
01 · Triggers

Four ways to wake an agent.

Most platforms ship with one trigger type. DeepAlpha gives agents four — the same invocation surface covers systematic rebalancing, scheduled research sweeps, one-shot event studies, and tape-speed breakout hunting.

Trigger spec
BTC/USDT ↑ cross 68,500

React the instant a level breaks — sub-second median dispatch from the tape.

Next firings
ARMED · now
agent.pypython · deepalpha/v2
1# wake on price
2from deepalpha import agent, market
3
4@agent.on_trigger(
5 market("BTC/USDT").cross_above(68_500)
6)
7def breakout(m):
8 return m.confirm_breakout()
02 · Model Stack

Five models. One consensus.

Every trade idea is voted by an ensemble that cross-validates gradient boosting, neural sequence models, and graph learning. Battle-tested in live markets.

Gradient boost
Best 7d Sharpe
LightGBM
BNB 7d · 68.1% win · Sharpe 1.56 · top performer on large caps.
Randomized trees
Lowest drawdown
ExtraTrees
XRP 7d · 63.6% win · Sharpe 2.92 · Max DD −3.0%.
Gradient boost
Baseline classifier
XGBoost
Direction classifier across all 21 assets at 4d and 7d horizons.
Categorical boost
Regime-aware
CatBoost
Encodes macro regime and whale-flow categorical features.
Bagging ensemble
Consensus anchor
RandomForest
Conservative vote — pulls consensus toward hold in low-info regimes.
Walk-forward validated across 21 crypto assets with 200+ engineered indicators per asset. Every prediction passes a ≥3-of-5 consensus vote before sizing — the highest-confidence band (AUC >0.85) has historically delivered a 84.3% hit rate and +5.15% average trade return.
21
Assets
200+
Indicators
≥3/5
Consensus
03 · Agent Spawn

Agents that never sleep.

Spawn autonomous agents on intervals, cron triggers, or macro events. Each agent routes to the five-model ensemble for entry, exit, and risk decisions — then acts on real-time price action.

3 Active Agents
XGBoost ensemble · 4H
Healthy
Day Trader
Every 1h · 24/7·8m ago
Volatility Scanner
Every 2h·22m ago
Breakout Hunter
Every 3h·1h ago
Macro Playbook
Sun 6:00 PM·2d ago
Live feed4H · 7H
BTC/USD
Consensus: BUY
97,150
+0.000%
ETH/USD
Consensus: BUY
3,420
+0.000%
SOL/USD
Consensus: BUY
192.4
+0.000%

Model votes: XGBoost ✓ LightGBM ✓ DNN ✓ LSTM ✓ GNN ✓5/5 consensus — spawning trade

04 · How it works

From idea to filled order in four steps.

Each trigger invokes an agent that reasons with the ensemble, attaches the right skills, and routes orders — all with an audit trail and a kill-switch at the broker.

01
Describe the thesis
A natural-language brief or a Python handler.
NL or code
02
Choose a trigger
Interval, cron, timestamp, or price action.
4 modes
03
Attach skills
Market data, ensemble ML, broker connectors.
40+ skills
04
Agent trades
Orders route to your broker with audit trail and kill-switch.
Live execution
05 · Strategy & Coverage

Full-stack market read.

A unified view of technicals, macro, and flow — translated into tradeable decisions at every bar.

20+
Assets covered
4H · 7H
Prediction horizons
24/7
Market coverage
5-model
Ensemble

Signal Inputs

  • Ichimoku Cloud
  • Bollinger Bands
  • RSI divergence
  • MACD
  • FOMC impact analysis
  • Volatility squeeze
  • Momentum breakout
  • Cross-asset correlation
06 · Track Record

Track record, on request.

Performance metrics, attribution, and drawdown analysis are disclosed to qualified investors upon request.

Request the deck
Equity curve — illustrative2024 — 2026

Disclosed privately to qualified investors. Contact investor relations for the full deck.

DeepAlpha — AI/ML Hedge Fund