Ask financial questions that need data and code.

Two things make it work: a hosted lake of historical market data, and research agents that answer by writing and running code over it. Built on the same Rebyte infrastructure as the rest of your team.

01 · Hosted data

Institutional-grade market data, already loaded.

US equities and China A-shares — from minute-level prices to full company financials and news. Nothing to license, download, or clean: the data is already here, consistent, and ready to query.

US 1-minute bars · 2021–2026 · ~2B rowsA-share daily history back to 1990SEC fundamentals, every line itemNews with semantic searchRefreshed daily

US Markets

us.bars_1mIntraday minute barsPer-stock, per-minute open, high, low, close, volume, and trade count — five years of it, about two billion rows. The granularity you need for intraday signals, execution studies, and precise backtests.

us.eodEnd-of-day pricesOne clean daily record per stock — open, high, low, close, volume, and trade count. The backbone for daily strategies, charting, and long-horizon studies.

us.fundamentalsCompany financialsOne row per company per fiscal period, straight from SEC filings: income statement, balance sheet, cash flow, comprehensive income — hundreds of line items, normalized across companies.

us.newsNews, searchable by meaningHeadline, full text, timestamp, and the tickers each story is about — every article pre-indexed by an AI embedding, so you can search by meaning, not just keywords.

China A-Shares

cn.bars_dayDaily bars, adjustedPer-stock daily prices with change, turnover, and adjustment factors — going all the way back to 1990.

cn.bars_1mIntraday minute barsPer-minute OHLC, volume, and turnover across the A-share market.

cn.daily_basicDaily valuation metricsPE and PE (TTM), PB, PS, turnover rate, dividend yield, shares outstanding, market cap — one snapshot per stock per day.

cn.incomeIncome statementFull quarterly income statement — revenue, operating and net profit, EPS, R&D, EBIT, EBITDA — with announcement and report-period dates.

cn.balancesheetBalance sheetTotal assets, liabilities, and equity, broken out across every standard line item.

cn.cashflowCash-flow statementOperating, investing, and financing cash flows with the net change in cash and its components.

cn.fina_indicatorFinancial ratios & indicatorsPre-computed profitability, margin, growth, leverage, and per-share metrics.

Every dataset shares a consistent schema and a stable primary key — no duplicates, no drift — queryable in SQL, with semantic search over news, and refreshed automatically every day.

02 · Answer with code

Your question becomes code. The code becomes evidence.

The agent writes and runs real code against the lake — joins, event studies, regressions — and returns charts, tables, and the code itself for review. The flagship run: a backtest.

Event study

"Is market volatility elevated in the week before every Fed meeting — and does it revert after?"

Take every FOMC date since 2015, compute realized volatility from minute bars in windows around each meeting, control for earnings season and macro releases, and test pre/post differences for significance.

Specific

"Did NVDA keep drifting after earnings beats since 2021?"

Run an event study with earnings dates, expected surprise, abnormal returns, liquidity buckets, sector controls, and post-event windows.

Modeling

"Does dividend yield predict 6-month returns after controlling for sector and market cap?"

Join point-in-time fundamentals, dividends, corporate actions, and price history, then test forward returns by quantile.

Backtest

"Buy AAPL when 20-day realized volatility falls below its 1-year 20th percentile and news sentiment is positive; exit after 5 trading days. Verify it."

Translate the rule into code, define the universe and costs, run the historical test, inspect failure cases, and return charts, tables, and caveats.

Every backtest, from hypothesis to evidence pack:

1Define signal, universe, rebalance schedule, benchmark, fees, and slippage.

2Build point-in-time joins across prices, fundamentals, dividends, news, and events.

3Run the backtest, sensitivity checks, ablations, and failure-case analysis.

4Return code, data lineage, charts, metric tables, assumptions, and review notes.

Ask questions like a professional research desk.

Start with "is volatility elevated before Fed meetings?", paste a strategy rule, or ask for a full backtest.