Kruncher Logo
Knowledge Infrastructure · How it works

You cannot prompt your way out of fragmented data.

The reason the modules can be trusted: one resolved, current, thesis-aware view of every company, person, and fund your firm touches.

One knowledge graph: companies, people, and funds resolved into a single connected record.

The problem

Six reasons raw data breaks AI

Foundation models solved reasoning. They did not solve the data. Six structural problems sit underneath almost every wrong answer.

Fragmented

One company shows up as five scattered versions across your tools.

Inconsistent

Sources disagree, and AI cannot tell which one should govern.

Incomplete

AI fills gaps by generalizing from public patterns, and passes guesses off as facts.

Changes over time

AI repeats an old number as if it were current, with no sense of history.

Needs standardization

Analysis needs a taxonomy and a peer set that is never actually defined.

Your view is missing

AI applies a generic lens, not your firm’s thesis and judgment.

Keynote · SuperReturn
Same Brain, Different Memory: Why your data, not your model, is the tech moat from here on.
Presented by Francesco De Liva, CEO at Kruncher
Watch on YouTube
How it works

From fragmented inputs to a record you can trust

Built bottom-up, the way the product does: from unstructured inputs to a resolved, standardized record with your judgment on top.

Sources from every role — documents, email, LinkedIn, and premium databases
Fragmented → Unified

One company, not five scattered versions

Every role in the firm captures data somewhere different. Kruncher ingests all of it, from CRM, documents, transcripts, email, messaging, LinkedIn and your address book, plus 20-plus premium sources, in any language, and resolves it into one clean record linked in a knowledge graph of companies, people, and funds.

A continuously running engine

Raw data and documents in, IC-ready reports out

Kruncher ingests everything, standardizes it, and keeps a living profile of every company, continuously. The same pipeline runs behind screening, diligence, and monitoring, so every module reads from the same current record.

  • Ingest all three layers: external data, your data rooms, your team’s decisions
  • Standardize: unstructured to structured, reconciled and normalized
  • A living profile: a time-series model, not a one-off snapshot
A continuously running data engine: ingestion, standardization, and a living company profile
Under the hood

The 14-step ingestion process

Every document and data feed passes through the same pipeline before it becomes a data point you can trust.

01
Connect your sources
02
Ingest documents & feeds
03
Parse PDFs, decks & Excel
04
OCR & text extraction
05
Detect language & translate
06
Extract entities
07
Resolve entities
08
De-duplicate records
09
Apply source precedence
10
Normalize fields
11
Standardize taxonomy
12
Enrich from 20+ sources
13
Build the knowledge graph
14
Update time-series & score
Global by default

Reads the world in its own language

Kruncher ingests 20-plus premium databases and public registries across the EU and beyond, plus documents in any language. Everything is normalized into structured English, with a link back to the original-language source, so your coverage never stops at the border.

  • Multilingual ingestion, normalized to English
  • EU and global public-registry coverage
  • The original-language source is always linked
Sources & registries
CrunchbasePitchBookCompanies House (UK)Cerved (IT)Bundesanzeiger (DE)LinkedInStatistaSemrush+ your CRM & documents
01Fragmented

Unified Company Record

Ingest everything and resolve it into one clean record, linked in a knowledge graph of companies, people, and funds.

02Inconsistent

Source Precedence

When sources disagree, the latest and most authoritative one governs. Contradictions are resolved, not averaged.

03Incomplete

Layered Record & Gap Detection

Labels each value by type (core data, premium, Kruncher estimate, your notes) and flags missing data instead of filling it. You always know a fact from an estimate.

04Changes over time

Living Company Profile

A time-series model and the Time Machine: a full change history where every change carries its source and date. No stale number repeated as current.

05Needs standardization

Standardization

Unstructured to structured: roughly 1,000 normalized data points per company, with consistent peer sets so companies are actually comparable.

06Your view is missing

Your Lens (Adaptive Learning)

Thesis, deal score, and preference learning sit on top. A thumbs-up teaches the system what your firm values, so later analysis reflects how you decide, never from shared-model training.

Specialists on the job
Deck readerTranscript analystFinancialsMarket sizingCompetitor scanTeam & foundersTraction signalsRisk flags
Not one model, many specialists

30+ specialized agents, working in parallel

Kruncher is not a single prompt. Dozens of specialized agents each read one slice, decks, transcripts, filings, competitor sites, and extract traction, unit economics, team and category signals, then hand off to the next. Purpose-built for private markets, not a generic LLM wrapper.

Layered record · Your lens on top

Public data at the base, your judgment at the peak

Every analysis is built in layers. Public data forms the foundation, premium sources and your internal research sit above it, confidential data from the target company adds depth, your thesis focuses it, and your judgment sits on top. Each value is labelled by which layer it came from, so nothing is a guess dressed as a fact.

A pyramid from public data up through premium sources, confidential data, your thesis, and judgment
Company analysis

Every company, fully analyzed

One standardized report, built the same way every time so companies are comparable.

MarketTeamCompetitive landscapeBusiness modelTractionFinancialsRisks
Your investment criteria

Scored the way your firm decides

Kruncher applies your thesis and up to 500 configurable signals, plus any you add, then returns a clear fit.

✓ Match, Deal Score 89/100✕ No match
A living company profile

Track how a company changes over time

Every data point is stored as a time series, not overwritten. You see how a company evolves quarter over quarter, when each number changed, and what it was before, so you understand trajectory, not just today’s snapshot.

Q1Q2Q3Q4Q5Now
Headcount, revenue, and every metric, tracked over time
Layered by design

Five layers of data on one record

Kruncher holds far more than public facts. Each layer is labelled, so you always know where a value came from.

Your firm’s opinion on the companyMost specific to you
Your firm’s knowledge
Company dataProtected by NDA
Premium data
External news & signalsPublic foundation
From documents to a knowledge graph

Not just data points. Relationships.

Kruncher turns every document into a knowledge graph that tracks companies, people, and investors, and the relationships between them. Those connections are what let it build accurate answers, not guesses assembled from scattered text.

CompetitorInvestmentStrategic buyer
CompanyPeopleInvestors
Kruncher captures your inputs, learns your rules and feedback, and activates outputs customized for your firm
Adaptive learning

A report specific to your firm, not the same one everyone gets

Kruncher learns from your past choices. Every IC decision, portfolio update, and piece of feedback sharpens your scoring, red-flag rules, and deal criteria, so the analysis reflects how your firm actually invests and gets sharper the longer you use it.

Where the advantage lives

Everybody rents the same model

The model is a commodity. Your edge is the knowledge layer beneath it: unstructured data turned structured, entities resolved, everything standardized, and the relationships and intelligence that only your firm accumulates. That layer is what Kruncher builds and keeps for you.

Everybody rents the same model; the knowledge infrastructure beneath it is the advantage
Your edge compounds

The longer you run on Kruncher, the sharper it gets

Every override, saved view, and diligence note sharpens the proprietary intelligence inside your isolated tenant. Your firm’s pattern recognition, what makes a great founder, what predicts breakout traction, what kills a deal at IC, gets encoded over time. No shared learning across customers; your edge stays yours.

What stays yours
Your data. Your tenant. Your model of the market.

Proprietary interaction data accumulates only for you, never used to train shared models.

Runs underneath everything

No black box

Follow any number in a report back to the document it came from. Here is how one data point gets built, step by step.

The pipeline
A source document arrives
1Extractionpull the raw facts
2Normalizationstandardize the format
3Enrichmentbenchmark vs external data
4Calculationderive the metric
The result, with its evidence
Source [1]

The documents highlight trends such as automation, modular APIs, embedded finance, and bank-fintech partnerships.

Extracted infoDocument1.pdf
Timeframe standardizedKruncher model
External benchmarkingCrunchbase
CAGR calculated using standard formula

Click any number to open its source.

Every data point carries a timestamp and the document, filing, or URL behind it. Every claim opens back to its source.

Trust & Control

Isolated Azure tenant, on-prem option, and full JSON export. Never used to train shared models.

ISO 27001SOC 2 Type IIGDPRAzure isolated tenant

A CRM, built in

Kruncher is also the CRM where you track investors, people, and companies, and the deal flow between them, so your relationships live on the same record as the intelligence.

InvestorsPeopleCompaniesDeal flow
For developers & platforms

The same layer, available as an API

Everything here is accessible programmatically. Call one primitive, analyze a company or document against a thesis, and get structured, sourced, timestamped output. Connect Claude, ChatGPT, or your own agents through the MCP server, or query the knowledge graph directly.

Explore Build on Kruncher →
Build on the primitives
Company Enrichment API
People Enrichment API
MCP Server
Knowledge Graph Access
Explore the building blocks

What the engine runs on

Every output is traceable, sourced, and audit-ready

Kruncher outputs are evidence-linked and governed, so your team can trust them and use them consistently. This is the layer every module reads from.

450
configurable signals
20+
premium sources
~1,000
data points / company
30+
AI agents