Research platform · AI agent
From an idea to a backtest report: factor library, single-factor evaluation, signal synthesis and backtesting joined into one pipeline, with an AI researcher that does the work.
One complete research pipeline
Four modules connect end to end: each step's output becomes the next step's input, and the agent can take over at any point.
- 01
Factor library
A large set of built-in classic factors, plus your own; pick a universe and dates and generate factor values in one click.
- 02
Single-factor evaluation
IC, Rank IC, ICIR, quantile returns and long-short NAV, net of fees and slippage.
- 03
Signal synthesis
Combine many factors into one signal: equal-weight ranks, rolling Rank IC weights, or your own Python.
- 04
Backtesting
Turn signals into positions: Top K long, quantile long-short or custom weights, with every cost visible.
The AI agent runs through it all: it understands your question, calls the right module and hands the results back in order.
Describe the idea. The agent does the work.
It does more than answer questions. Aleph Agent frames research questions, analyses data, builds a research plan, and calls single-factor evaluation, signal synthesis and backtesting directly, bringing back the results of every step.
Frame the question
Turns a vague idea into testable hypotheses, with clear data, periods and criteria.
Analyse the data
Reads your factor and result files and points out anomalies, gaps and what is worth a closer look.
Build a research plan
Lays out the steps and control experiments so every conclusion is backed up.
Do the work
Calls evaluation, synthesis and backtesting with your parameters, then explains the results.
Factor library: from classic factors to your own ideas
- Built-in factors organised by category: volatility, trend, liquidity, reversal, price-volume and momentum, with advanced factors added over time.
- Write your own factors in Python with one input/output format, and use them alongside the built-ins straight away.
- Choose a universe and date range, generate a factor-value package in batch, and use it in every later module.
Single-factor evaluation: is this factor any good?
- IC, Rank IC, ICIR and coverage show a factor's predictive power and stability at a glance.
- 5- or 10-quantile backtests and long-short NAV show whether returns rise steadily across quantiles.
- Costs are deducted using per-side fees and slippage, with turnover reported; external factor files (CSV / Parquet) are supported too.
Signal synthesis: make many factors speak with one voice
- Equal-weight ranks: average each factor's daily cross-sectional rank. No model training, robust and transparent.
- Rolling Rank IC weights: recent performance sets each factor's direction and weight.
- Custom Python: write your own logic in a sandbox that only sees data known at the time.
Backtesting: turn signals into positions you can test
- Top K equal-weight long, quantile long-short, or custom weights in Python.
- Daily holdings, trades and costs, with NAV and drawdown shown against a benchmark.
- Every backtest records its parameters and data, so results can always be reproduced.
Non-negotiables for serious research
Private by default
Factor code and data you upload are visible only to you.
Point-in-time
Signals open positions on the next trading day; backtests never use data that was not yet known.
Reproducible
Every evaluation and backtest is saved as an immutable result, with its parameters and data snapshot.
Whole-market coverage
Research across all A-shares or a chosen index universe, over any date range.
Want to see what Aleph Agent can do for your research?
Write to us and we will get back to you soon.
info@alephedge.com