HedgeIQ

(HedgeFund Data Intelligence Accelerator)

The AI Workbench for Hedge Fund Investment Teams

Book a Demo
See the agents

Runs inside your Databricks estate. No data export. Decision support only.

Specialized AI agents for long/short equity funds, built natively on Databricks. Portfolio managers ask in plain English and get cited answers from governed fund data, internal research and market context.

Market data shows you what happened. HedgeIQ shows you what it means for your fund.

A Portfolio Manager's Day Is Split Across Too Many Places

Every answer means stitching reports together by hand, and the most important risk often isn’t written down anywhere.

  • Commissionsin one system
  • Fund performancein another
  • Researchin folders and inboxes
  • Market newsin the terminal
  • Thesis riskin people's heads

That Creates Three Problems

Answers Take Too Long

A routine desk question becomes a fifteen-minute scavenger hunt through broker statements and spreadsheet reconciliations.

Monitoring Is Inconsistent

Theses are tracked through email reminders and memory, so broken assumptions surface late.

Decisions Are Hard to Audit

When the investment committee or compliance asks what was known and when, there's no single record.

One Screen for the Entire Investment Desk

HedgeIQ brings the core workflows of an investment desk together, with specialized agents reading governed data you already own.

Instead of jumping between dashboards, spreadsheets and research systems, PMs and analysts ask one question and get one unified answer.

These are purpose-built agents for specific desk workflows, not a generic chatbot.

  • 01

    Ask Once

    Commissions, fund performance, research and meeting notes come back as one answer.

  • 02

    See the Source

    Every answer carries its source table or document and its refresh date.

  • 03

    The PM Decides

    Agents gather and monitor. Buy, sell and sizing decisions stay with the desk.

Three Agents, One Desk

Each agent handles a distinct part of the investment workflow, from the first question to ongoing monitoring of a live position.

Fund Intelligence Copilot

Plain-English answers over your fund data

Ask about commissions, broker activity, fund performance, holdings, meeting notes and research documents. The copilot translates the question into queries against governed tables and documents, then returns a ranked, visual answer in seconds.

Table-level source citations mean the investment committee and compliance can trust what they're reading.

Questions the desk asks

Which brokers have the largest commission delta?
Show fund performance for June 2026
What drove sector attribution last month?

Investment Research Agent

Structured research workflows, same quality every time

Slash commands replace open-ended chat, so every analyst gets the same structured output. The agent assembles a governed fact pack first (market data, consensus, holdings, meetings) and the model drafts the narrative second.

Fifteen minutes of prep becomes one standard card the whole desk reads the same way.

Research Lab commands

/scanCompany snapshot in about 30 seconds: price, consensus, book context, meetings and news
/catalystTurns an idea into a sourced brief with thesis pillars, measurable monitoring triggers and dated catalysts
/memoIC-ready investment memo
/briefMorning brief handoff for the desk

Thesis Sentinel

The control tower for active views on the book

Sentinel monitors every active thesis and scores its health daily against news, filings and alternative data. Each trigger is marked confirming, weakening or breaking, with AI reasoning layered on governed signal data.

When assumptions break, it raises an alert. Each morning it produces a portfolio brief across monitored names and open alerts.

It doesn't try to predict the market. It operationalizes discipline, so the desk sees risk before it becomes a surprise.

Company Name · Long72 / 100

Unit economics inflecting as drive-through lanes lift throughput at lower build cost

ConfirmingSame-store traffic above consensus
ConfirmingNew-unit build cost trending lower
WeakeningRestaurant-level margin versus threshold
BreakingPace of new drive-through openings

From an Idea to a Monitored Thesis

The agents connect into one workflow, so a conversation on the desk ends up as a tracked position with clear tripwires.

HedgeIQ — How it works

  1. Ask

    A PM asks the copilot a fund question and gets a cited answer from governed tables.

  2. Research

    Run /scan on a name for a standard research card built from a governed fact pack.

  3. Formalize

    Run /catalyst to turn a verbal idea into pillars, measurable triggers and catalysts.

  4. Activate

    One click writes the thesis to the registry in Unity Catalog and turns on monitoring.

  5. Monitor

    Sentinel scores thesis health daily, raises alerts and feeds the Morning Brief.

Built on the Databricks Estate You Already Trust

HedgeIQ runs as a Databricks App, inside your existing controls. There's no new infrastructure and no data exported to a vendor platform.

Unity Catalog
Governed Delta tables, thesis registry and ops log
Genie
Natural language to SQL over fund and commission data
Vector Search
Retrieval across research documents and meeting notes
Model Serving
Configurable model endpoint per agent
LangGraph agents
Orchestration of specialised desk workflows
MLflow
Full trace behind every agent call
SQL Warehouse
Direct queries against governed warehouse data
PM-Scoped Access
SSO plus portfolio scoping. Each user only sees the tickers they're entitled to see.
A Full Audit Trail
Every run is logged: who asked, when, which agent, tools called, data returned, latency and outcome. Compliance can answer what the system said on any given date.
Configurable Without a Code Release
System prompts, guardrails and model endpoints are managed in Settings. Prompt versions are stored for compliance review.
Decision Support Only
Agents never make buy, sell or sizing decisions. Interpretation is labelled separately from data.

What Changes for the Desk

Faster Decisions

Questions that took fifteen minutes return natural-language answers in seconds, over live governed tables and documents.

Higher Trust

Every material claim is sourced, and agent interpretation is always labelled separately from the underlying data.

Better Coverage

Daily thesis health scores and morning briefs reduce missed signals across the book.

Lower Risk

Data stays in Unity Catalog, access is scoped per PM, and the system supports decisions without making them.

Faster Time to Value

With fund and commission tables already in Databricks and a schema-aligned prototype, a pilot starts in weeks, not quarters.

A Reusable Platform

A shared tool layer of Genie, Vector Search, SQL and web evidence means new agents are assemblies, not rebuilds.

Watch the Walkthrough

Follow a portfolio manager from a plain-English question to a cited answer, a research brief and a monitored thesis.

What the Pilot Proves

The Databricks-native agent pattern has been demonstrated end to end on real desk workflows.

What the Pilot Proves

  • Fund Questions Become Cited Answers in SecondsGenie and SQL
  • Research Snapshots Are StandardizedGoverned fact packs
  • Theses Are Monitored With Explicit Break ConditionsThesis Sentinel
  • Every Run Leaves an Audit TrailOps log and MLflow

Path to Production

  • Broader ticker coverage
  • Live market data feeds
  • Firm-wide role-based access
  • Training for the desk
On the Roadmap: Research Verifier

An agent that checks sell-side claims against governed data. It is not part of the current pilot scope.

Frequently Asked Questions

AI agents like HedgeIQ let portfolio managers ask questions in plain English and get cited answers from governed fund data in seconds. Questions about commissions, broker activity, fund performance or holdings no longer mean a fifteen-minute hunt through spreadsheets and reports. The agent gathers the data, and the PM makes the decision.

An agentic AI workbench is a set of specialized AI agents built for specific investment workflows, rather than a general-purpose chatbot. HedgeIQ includes a Fund Intelligence Copilot for data questions, an Investment Research Agent for structured company research, and Thesis Sentinel for monitoring active positions. All three work from the same governed data.

Yes. HedgeIQ runs natively as a Databricks App, using Unity Catalog, Model Serving, Genie, Vector Search and MLflow. Fund data never leaves your Databricks environment or goes to a third-party vendor. Access is controlled through SSO and portfolio scoping, so each portfolio manager sees only the tickers they are entitled to see.

 

HedgeIQ's Thesis Sentinel turns each investment thesis into measurable monitoring triggers. It then scores thesis health daily against news, filings and alternative data. Each trigger is marked confirming, weakening or breaking. Alerts go out when key assumptions fail, and a daily Morning Brief summarizes risk across the book.

With HedgeIQ, a portfolio manager can ask a question like “Which brokers have the largest commission delta?” and get a ranked chart in seconds. The answer is generated from governed commission tables through Databricks Genie. Each result includes table-level source citations, so compliance and the investment committee can verify the figures.

For firms whose fund and commission data already sits in Databricks, a HedgeIQ pilot can start in weeks rather than quarters. The accelerator uses a schema-aligned prototype and a reusable tool layer, so no new infrastructure is needed. Additional agents can be assembled from existing components instead of built from scratch.

See HedgeIQ on Your Own Fund Data

Book a walkthrough with Kadell Labs and scope a pilot on your Databricks estate.

Book a demo