Research that used to take days, and dashboards that finally settle the argument
We build data platforms, analytics and Claude-based AI for trading platforms, fintechs and the finance teams inside them. Two pieces of our published work are in this industry: an AI layer at StockJarvis that cut research and documentation time by 60 to 70%, and a Series C fintech that went from 240 dashboards to 8 and saved $72k in its first year.
- Published work in this industry
- 1 customer story, 1 case study
- Research and documentation time cut, StockJarvis
- 60 to 70%
- Saved in year one, Series C fintech
- $72k
- Team on a build
- Pod of three, no bench
On this pageWhat we hear in financial services
What we hear in financial services
Each of these comes from our published work in financial services or from an article we wrote for teams in it.
- 01Analysts spend their days writing up results
At StockJarvis, analysts spent significant time reviewing multi-leg backtest outputs, interpreting Greeks and payoff profiles, and writing strategy summaries by hand.
- 02240 dashboards and still no answer
A board was built for every question anyone asked, on a six-figure licence. Opening one rarely ended the argument, so people went back to asking an analyst.
- 03Customers want a login to your dashboards
The charts already exist, so it looks like a week of work. What changes when customers log in is isolation, load shape and a support queue.
- 04The assistant said a claim was covered. It was not.
When an AI feature goes wrong, only the record of each model call can tell you whether the model invented the answer or was handed the wrong document.
- 05Data that is not allowed to leave the country
Picking a region satisfies the questionnaire. It does not answer where the backups, the logs and the vendor's support tooling actually sit.
- 06The real finance model is a spreadsheet
The workbook that decides the hiring plan and what gets said to the board is the system of record, and nothing in the warehouse comes close to it in influence.
What we have delivered
Two published pieces of work. StockJarvis is named; the fintech case study is anonymised.
StockJarvis, an AI layer built on Claude inside their Strategy Lab, Simulator and Algo modules.
The platform the layer runs inside, live since early September 2026.
Eight boards built new around the decisions the business makes.
A six-figure BI licence replaced with a tuned embedded Superset stack.
Read the full accounts
- Customer storyStockJarvis: an AI layer inside India's precision trading platform
Built on Claude by a woodfrog pod of three in four months, and used by the research and product teams every day.
Read the StockJarvis story ↗ - Case study, fintech240 dashboards, one six-figure licence, and no answers
Six weeks, fixed scope and fixed price. Load time from 15 seconds to 3.
Read the case study ↗
Where to start
Four places a first engagement can begin. Week 1 tells you which one comes first.
- AIAn AI layer over your own research and documents
One agent on your live data, with a kill switch, built directly on Claude. The same way of working that runs at StockJarvis.
AI agents ↗ - AnalyticsFewer dashboards, on Apache Superset
Off a licensed BI tool, keeping only the boards that carry a decision. Built and run by contributors to the project.
Apache Superset ↗ - DataOne number finance and sales both accept
One set of records both teams accept, with the definitions agreed once.
Data integration and governance ↗ - AIAI that holds up in review
Graded evaluation sets before launch, and governance that is defensible in a board meeting and auditable on demand.
AI governance ↗
How the first weeks run
The same three phases in every industry. What changes is what we build in them.
- Week 1, fixed feeAudit
Two calls and a written one-pager. Yours to keep whether or not you continue.
- Weeks 2 to 6Build
Something real in front of real users by week three. Median six weeks to the first production-grade artefact.
- Week 7 onwardOperate
Quarterly reviews and on-call governance. The people who built it pick up the phone.
Questions we get asked
Further reading
- Logging every model call is the cheapest governance you will buyWhen an AI feature goes wrong, only the call record can tell you what happened.
- Data that is not allowed to leave the country changes the architecturePicking a region satisfies the questionnaire. It does not answer where the backups sit.
- Putting dashboards in front of customers turns reporting into a productThe charts are the easy part.
Start with the Week 1 audit
Two calls and a written one-pager naming the first thing worth building, and who owns it. You keep it either way. NDA-friendly, fixed scope. Write to hello@woodfrog.tech.
