Sparkworks Book a data therapy consultBook a consult

AI made a mess.
We put on a clinic.

Book a data therapy consult
Connecting the dots · enabling new forms of access

For teams whose AI got ahead of their data.

Our forward-deployed team uncovers root problems, provides end-to-end fixes, and leaves you with AI that holds up in front of clinicians, scientists, customers, regulators, and your board.

CLEAN UP

Fix what you already have

We audit the code, models, prompts, pipelines, and vendor licenses in place today, then put back the lineage, testing, and observability that make them trustworthy and enterprise-grade.

BUILD & DEPLOY

Ship new products safely

We design and build AI products inside your existing controls. High governance and regulatory compliance are part of the plan from the first commit, on whatever cloud and models you already use.

TRANSFORM

Bring your people along

AI only works when the organization changes with it. We educate and train your team and help evolve your policies and processes so new systems get adopted, not just installed.

HOW AN ENGAGEMENT RUNS
01
Intake
What you have built, what is stuck, and what a 10/10 outcome looks like.
02
Diagnosis
A hands-on review of your tools, models, processes, policies, and roles.
03
Treatment
We fix, rebuild, or ship alongside your team. Not just pieces of code, but the holistic system.
04
Follow-up
Monitoring, maintenance, and improvements to ensure results are impactful and sustainable.

Our healthcare data work, in more detail.

Three projects, written up plainly. Client names and exact figures stay private. Where we show numbers, they come from tests on synthetic data or are marked approximate.

HEALTHCARE DATA · CASE STUDY

Teaching a data agent the rules of the data.

A life-sciences team wanted researchers to ask questions about real-world oncology data in plain English and get patient counts back. The agent worked, mostly. The trouble was that its wrong answers looked exactly like its right ones.

DOMAINReal-world data · HEOR
OUR ROLEDesign and build
DATALicensed oncology data products, inside the client’s own warehouse
STATUSAgent built; safeguards tested on synthetic data
2 → 16
Questions right out of 16 once the agent had checked rules, not just column names
14
Wrong answers in the first round that came back with no error at all
yes ≠ True
One mismatch between a vendor manual and the data made whole queries return zero
600
Synthetic patients and 16 questions, with every right answer known in advance
TRY IT · BREAK THE AGENT, THEN FIX IT

Pick a question, then change what the agent is given. These are the real results from our test on synthetic data.

{{av.q}}
AGENT ANSWERED{{av.said}}
CORRECT ANSWER{{av.truth}}
VERDICT{{av.verdict}}

{{av.why}}

WHAT WE DID
01
Read the sources

Each licensed data product comes with a data dictionary and a usage manual. We read both and turned them into short, single-statement rules, each pointing back to its page.

02
Test every rule against the data

Before a person approves a rule, it is run against the real tables. Do the columns exist? Do the values match what is stored? Is it really one row per patient? The manual and the dictionary disagreed on values like yes and True, so this step mattered.

03
Give the agent only what is approved

For each question the agent gets the approved rules and the real column values for the tables it needs, and nothing else. A query that returns zero rows goes back as a failure, not as an answer.

The agent could always write SQL. What it lacked was the fine print of the data. Once it had that, 16 of 16 test questions came back right, with no retries.
AI EVALUATION · CASE STUDY

Can a small decision model check the agent’s work?

Most AI writes text. Jev, a new model from TypeSafe AI, picks from options we give it, like yes or no, and says how sure it is. We tested whether it could sit in front of our healthcare-data agent and check each plan before any number reaches a user.

MODELJev, by TypeSafe AI
OUR ROLEEvaluation and design
TEST DATASynthetic only
STATUSExperiment. Nothing is in production.
14 / 14
Planted mistakes caught in oncology questions, with 1 false alarm in 16 correct plans
≈ 0.3 s
To check one plan, against about 3 seconds for a general-purpose AI
≈ 50¢
For ten thousand checks
2 → 11
Questions right out of 16 when Jev checked plans against the rules, up from 2 with a dictionary alone
PLAY · CATCH THE SLIP

Six real plans from our tests. Some are right, some quietly answer the wrong question. Read the plan, make your call, then see what Jev said.

{{jTag}}
{{jq.q}}
THE AGENT’S PLAN
{{p}}
{{jr.stamp}} {{jr.you}}

{{jr.why}}

How sure Jev was the plan fits the question{{jr.conf}}
{{jr.said}}what the report would have said
{{jr.right}}the right answer
FINAL SCORE
{{jScore}}
You
6/6
Jev, in {{jJevTime}} in total

{{jEnd}}
WHAT WE LEARNED
01
It enforces the rules it is given

Jev checks a plan against a rulebook. If the rulebook is wrong, so is the check. With a checked rulebook the agent scored 16 of 16 on its own, so the rulebook comes first and Jev is the second layer.

02
It decides, it does not calculate

Jev does not count, do date math or write reports. The agent and the database still do that. Jev only judges whether the plan fits the question.

03
The free option was not ready

We also tested an open-source look-alike that runs on our own machines. It was fast, but out of the box it was worse than guessing on three of seven tasks, so we parked it.

A cheap second pair of eyes for the agent, as long as the rulebook is right. That is where we would start.
KNOWLEDGE GRAPH · CASE STUDY

One graph across the warehouse and the papers.

The client’s structured data lived in the warehouse. What people actually knew about a disease, a drug or a biomarker lived in internal publications and reports. Nothing connected the two. We built a knowledge graph over both, so a question about one thing leads to everything linked to it.

DOMAINKnowledge graph · discovery
OUR ROLEDesign and build
SOURCESData warehouse tables and internal publications
STATUSIn production
~30
Warehouse tables mapped into the graph
~1,500
Internal publications and reports linked in
~10K
Entities such as diseases, drugs, biomarkers and trials
4
Standard vocabularies used to line up names: ICD-10, SNOMED CT, LOINC, RxNorm

FIGURES ARE APPROXIMATE AND ROUNDED. EXACT COUNTS AND NAMES ARE WITHHELD TO PROTECT THE CLIENT.

TRY IT · FOLLOW THE LINKS

A simplified picture of how it connects. Click any node, or start from a question. Teal nodes are entities, gold are warehouse tables, purple are publications.

{{gsel.kind}}
{{gsel.label}}

{{gsel.note}}

DIRECTLY LINKED TO {{gsel.links}}
HOW IT WAS BUILT
01
Start from the questions

We began with the questions people actually ask, then defined a small set of entity types to answer them: diseases, biomarkers, therapies, trials.

02
Link the sources

Warehouse columns are mapped to those entities, and the same entities are pulled out of the publications. Names are matched through standard vocabularies, so different spellings of one thing land on one node.

03
Keep the receipts

Every link records where it came from, a table and column or a page in a document, so any answer can be traced back and checked.

What used to mean asking three people and opening a dozen files is now one lookup, with the sources attached.
{{csCount}}
{{c.tag}}
{{c.name}}
{{c.area}}

What we solve in healthcare and life sciences.

We help executives and their teams make faster, more precise decisions grounded in data. The foundations stay the same: connected knowledge, governed data, and evidence you can trace.

WHAT IT DOES

{{item.desc}}
Have a different problem? Book a data therapy consult

Sometimes the answer is something new.

We also embed with organizations to build something new: a new business, a new internal capability, a new bet. Same network, new connections.

NEW LINE OF BUSINESS

Start something adjacent

From thesis to first revenue. Venture design, product build, pilot contracts, and the operating model to run it, inside the enterprise.

LABS FUNCTION

Stand up an AI lab

A spun-out team with its own charter and governance that launches new AI initiatives without disrupting the core business.

NEW TECHNOLOGY

Experiment at speed

Two-to-six-week sprints on new technology, pointed at a real business question. You get working prototypes and a plan for scaling.

We build with

We do not sell a platform. We work in your stack, whatever it is, and connect the pieces so your data, models, and teams actually talk to each other.

We also build, operate, and advise new ventures.

Alongside client work, we help start and grow new companies in healthcare and life sciences.

BUILDProduct, platform, and data foundations, from zero to first customers.
OPERATEFounding and fractional leadership through launch and growth.
ADVISEStrategy, go-to-market, and technical diligence for founders and boards.
{{vCount}}

New forms of access outside the clinical setting.

Work across consumer and B2B lending and payments: companies we have started, built alongside, or advised.

CONSUMER LENDING · CASE STUDY

Standing up KMRF.

Kawasaki and ITOCHU needed a U.S. captive lender and had nothing to start from. No entity, no licenses, no systems, no team. We ran the build end to end, from formation and licensing through a national dealer launch, with CI/CD and AI-assisted development in place throughout.

CLIENTKawasaki Motors Retail Finance (KMRF)
BACKED BYKawasaki Motors Corp., U.S.A. & ITOCHU
OUR ROLEFounding Operators
ENGAGEMENTFeb 2025 – Present
50 + DC
States licensed for sales finance, in four waves
~4 mo
From kickoff to the first loan booked
80+
Dealers enrolled at rollout, with 200+ more in the pipeline
40+
Roles designed and hired into a company that started with no one
TIMELINE · BLANK PAGE TO NATIONAL LENDER
WHAT WAS BUILT
{{kc.n}}

{{kc.title}}

{{p}}
{{v}}

A licensed, staffed, working U.S. captive lender, built from a blank page to a national lending business in about a year, and still supported today with ongoing maintenance and enhancements.
{{fCount}}
{{c.name}}
{{c.area}}
{{c.badge}}
{{c.domain}}
LAST UPDATED 29 SEP 2026

What we are paying attention to right now.

Technology we are testing now, and where we think it lands in regulated industries.

{{nCount}}

How can we help?

Book a data therapy consult