Research-embedded · Financial markets

Agents you can trust, on data you can trace.

We build financial-market products for fragmented markets. Data first, then agents, with a person approving what matters.

Start a research projectSee what we have built
Our own operations run this way today. Products follow, one step at a time. The three steps
How we work

Every product starts as a research question.

What holds up becomes a product and runs live. What does not stays research, and we say so.

Our approach: data first, then agents
From research to where it is now
Options market analyticsRunning live, daily

Live tick data, market observation, recurring backtests · Digital assets

Hephaistos and AthenaPublished · used by SaaS clients

Unified order books, smart order routing · Digital assets

QAVCMAlpha · ZIM-funded

Carbon market data and valuation · Carbon

ImmoTokenResearch and concept

Tokenized real estate · Real-world assets

Projects

Three markets we have worked in, at different stages.

01Running live

Digital asset markets

Market data, order books and options analytics, live every day.

Options market analytics
Steps 1 and 2
More
  • Options market analytics: live tick data, market observation and recurring backtests
  • Hephaistos: unified order books across crypto exchanges
  • Athena: smart order routing on that order book
Analytics and backtesting run by agents, supervised (step 2). Research and reporting: a person approves every result (step 1).
02Alpha · seeking partners

Carbon markets

QAVCM: one traceable view of a scattered voluntary carbon market.

QAVCM project page
Step 1
More
  • Harmonised data on [11,659] carbon projects from several registries
  • Documented data lineage behind every figure
  • Explainable valuation and a procurement agent prototype
  • Research prototype, publicly funded (ZIM), with XU Exponential University
Step 1: AI-supported, a person approves every output.voluntarycarbon.org
03Research

Tokenized assets

ImmoToken: concepts for tokenized real estate in Germany.

ImmoToken project page
Research concept
More
  • Technical and regulatory concepts under Germany's Electronic Securities Act
  • Research project and book chapter (Haufe, 2024)
  • Available as a project on request
Concept stage, not a running product.immotoken.org
From AI support to agents

In three steps to your own agents.

People move from in the loop to on the loop, never out of it, one part at a time.

Here today: a person approves every resultHere today: agents run it, a person supervisesTarget
1 · TodayAI supports, a person approves
2 · NowAgents run parts, supervised
3 · TargetAgents run the routine
Research and reports
Analytics and backtesting
target
QAVCMCarbon market prototype [stage to confirm]
A person approves every result
next
target
ImmoTokenTokenized real estate, research project
Research concept, not running
Our products
Step 1 · TodayAI supports the work
Step 2 · Changing nowAgents run defined parts
Step 3 · TargetAgents run the routine
Research and reports
Analytics and backtesting
target
QAVCMCarbon market prototype [stage to confirm]
A person approves every result
next
target
ImmoTokenTokenized real estate, research project
Research concept, not running
A product built with youYour product walks the same three steps.
How a project runs
Your new productLive from day one. You set the pace.
A person approves every result
next
target
Your new productLive from day one. You set the pace.
A person approves every result
next
target
What moves a part up?Evidence decides when an agent takes over.
01
The data layer and its lineage are documented.
02
Agent results were accepted without change over an agreed period.
03
It is clear what a person approves, what they watch, and how problems are escalated.

Example conditions, agreed per product. More on the approach page

For banks and financial institutions

Find your next product, and build it on agents from day one.

No finished question needed. We find where agents can help, pick the product worth building and run it beside your existing business.

How a project runs
Your existing businessYour processes and systemsUnchanged while the new product proves itself.
we look at it with you
The new productRuns beside your business, not inside it.
  1. 01Find where agents can helpFinding the data is our job, not your homework.
  2. 02Choose the next productThe best opening becomes a research question.
  3. 03Build it agent-first, beside the oldLIVE FROM DAY ONE · STEP 1
    Own data layer and agents. A person approves every result.
  4. 04Move up, then carry it overFrom approval to supervision, once results hold.
what has proven itself
Back into your existing businessProven parts carried overOnly what has held, at your pace.
  1. 01Find where agents can help
  2. 02Choose the next product
  3. 03Build it agent-first, beside the old
  4. 04Move up, then carry it over
Your existing business
Your processes and systemsUnchanged while the new product proves itself.
Proven parts carried overOnly what has held, at your pace.
we look at it with you
The new product runs beside your business, not inside it.
what has proven itself
The new product
OpeningsWhere agents could help. Finding the data is our job.
One product chosenThe best opening becomes a research question.
LIVE FROM DAY ONE · STEP 1Built agent-firstOwn data layer and agents. A person approves every result.
STEP 1 → 2 → 3Moves up on evidenceFrom approval to supervision, once results hold.

Banks we work with are exploring carbon market products and agent-run process automation.

Option AUse a product we runFor example QAVCM. The IP stays with us.
Option BBuild a new product togetherThe new product belongs to you. [IP wording pending legal check]

Research projects can qualify for public co-funding such as ZIM. [confirm wording]

Research

Published, and next to the product it supports.

All publications
QAVCM · CarbonQAVCM: A Quantitative Aggregation and Analytics Platform for the Fragmented Voluntary Carbon MarketAKWI 2026, Berlin · acceptedAthena · Digital assetsAthena: Smart Order Routing on Centralized Crypto Exchanges Using a Unified Order BookInternational Journal of Network Management 34(4)Method · Agentic companyAlgorithmic Entrepreneurship: Data-Driven Algorithmic Automation and Output-to-Employee ScalingTH Wildau ENSP 3 (WiKKI26), 2026
Team

A lean senior core, with research talent behind it.

Experienced people lead every project. Researchers from HPI and XU Potsdam join where their work fits.

Robert HenkerCo-founder. Built a visual analytics research team at the Hasso Plattner Institute. [role line]
André HoldschickCo-founder. [role line and research affiliation]
Research benchResearchers and graduate talent from HPI and XU Potsdam.

Looking for your next product?

Tell us about your business. We reply with a first idea of where agents could help and what is worth researching.

[contact email]Domstrasse 12, Potsdam, Germany