OpenKuber · causal intelligence
Causal intelligence for Indian financial markets.
When an NSE stock moves sharply, IROHA separates what the market, the sector and the company each contributed, tests every proposed cause against the evidence, and turns the answer into stress tests and position sizing.
- Big RAYMOND moves since 2015 the model leaned the right way on
- 41 of 52
- Big RAYMOND moves since 2015 the model leaned the right way on
- Public shock scenarios with a known answer in IndiCause-Bench
- 1,532
- Public shock scenarios with a known answer in IndiCause-Bench
- Language models that lost money in LabhLabs Season 1
- 6 of 6
- Language models that lost money in LabhLabs Season 1
- Bank Nifty stocks covered by the API today
- 14
- Bank Nifty stocks covered by the API today
Core thesis
OpenKuber is building an AI-native market intelligence platform for Indian markets. Not a dashboard run by human analysts, but a system in which stateful AI agents observe live market data, reason over macro signals, institutional flows, corporate events and alternative datasets, and pass what they observe through a causal engine that produces causal chains and position sizing.
We deploy these agents against real NSE conditions to expose where technical strategies, stateless models and inherited US-market assumptions fail, then use those failures to build the causal engine itself: a citation-forward layer that learns which signals matter, why stocks move, and how a decision should be challenged.
India's messy, under-structured data is not a limitation. It is the opportunity.
01 · Causal engine
One stock’s causes, as a map.
Every territory is a variable IROHA uses for RAYMOND, from the market and industry peers to company news and NSE filings; the small cells are the data behind each. Arrows are causal links. Hover a territory for its evidence.
02 · Shock breakdown
Every RAYMOND shock since 2015, decomposed.
52 big moves, each split into what the model expected, the market, the sector and the company. The model leaned the right way before 41. Select a move to see where it came from.
RAYMOND · 52 big moves since 2015
call leaned right leaned wrong41/52
#52 · 21 Aug – 7 Oct 2026 · Company news
+52.8%move
Raymond Limited’s Aerospace Subsidiary Wins Multi-Programme Aerospace Orders from a Leading Indian Aerospace and Defence Major
Passed every causal test: policy news, news.
Where the move came from · pts
- Model call leaned the right way
- Market
- Textiles peers
- Company event
Call + market + sector + company = +52.8 pts. Data as of 27 October 2026.
03 · LabhLabs
Six language models. One month. No winners.
Each model traded a ₹1,00,000 paper portfolio of NSE stocks from 2 to 27 March 2026. The best, GPT-4.1 Nano, ended at -3.89%. Season 2 begins on 15 December 2026.
2–27 Mar 2026 · paper trading
04 · Developer API
The same answers, as data.
A headless causal API for institutions that build their own tools. Available to pilot customers.
{
"symbol": "NSE:RAYMOND",
"window": [
"2026-08-21",
"2026-10-07"
],
"move_pct": 52.8,
"components": {
"model_call": 2,
"market": -0.5,
"sector": -1.9,
"company": 53.1
},
"causes": [
{
"event": "Aerospace subsidiary wins orders",
"date": "2026-09-12",
"tests_passed": [
"policy",
"news"
]
}
]
}- 01
Explain a move
A symbol and a window in; a sourced decomposition into expectation, market, sector and company out.
- 02
Causal graph
The variables that drive a stock, with the direction and weight of each tested effect.
- 03
Scenarios
Specify a shock to any driver; receive a range of outcomes for each stock in the universe.
Live
Use the work, not just read about it.
Three public apps run on the same engine and data as this site. Each is built for a different reader.
raymond.openkuber.comShock Engine · RAYMONDEvery big move in one NSE stock since 2015, with headlines, filings, causal tests and a what-if engine.For · Portfolio managers and analysts
benchmark.openkuber.comIndiCause-BenchSeven models, 1,532 shocks with known answers, replayable regime by regime.For · ML and quant researchers- labhlabh.openkuber.comLabhLabs dashboardPositions, trades and chat logs of AI models paper-trading NSE stocks.For · Anyone curious how AI models trade
Research
Research, in the open.
We publish our methods and our results, including the ones that do not flatter us.
Who we work with
Research infrastructure for regulated firms.
Registered investment advisers, portfolio management services and proprietary and quantitative desks use IROHA to explain the moves in their books, with sources, and to stress-test them before the next one.
Engagements begin with a pilot of two to four weeks on the firm's own universe of stocks: a causal account of every large move, a stress-test report, and a review with the team.
Questions
Frequently asked
What does IROHA do?
IROHA explains why an Indian stock moved. It splits each large move into market, sector, macro and company-specific parts, tests each cause against evidence, and turns the result into stress tests and position-sizing inputs.
Is OpenKuber an investment adviser?
No. OpenKuber provides research infrastructure to regulated firms. It does not give buy or sell recommendations and does not serve retail investors.
Which markets does IROHA cover?
NSE-listed equities. The API currently covers the Bank Nifty constituents; wider universes are available in pilots.
