Kipi Technologies — AI Insights & Grades on Company Health

Why alternative data

Reported data is a snapshot. Company health is a moving picture.

Earnings, filings, and guidance are point-in-time snapshots — published quarterly, compressed into a handful of numbers, and describing a period that ended months earlier.

Meanwhile, companies emit a continuous stream of publicly observable operational evidence that never stops between reports. It's abundant — and it's also unstructured, noisy, fragmented across sources, and far too large for human analysis.

That's the problem Kipi Technologies was built to solve. Our models read alternative and traditional data together, continuously, and translate them into structured insights and company-level grades — a living picture of how each company is really doing.

Quarterly earnings report

~90 days behind

Analyst estimate revisions

~30–60 days behind

News & press coverage

Reactive

High-frequency alternative streams

Continuous

Unstructured public text at scale

Always current

Cross-source operational indicators

Refreshed weekly

The product

The Kipi Signal Platform

One platform, three layers: a multi-source data engine, a proprietary AI modeling core, and insight delivery built for analytical workflows.

Multi-Source Data Engine

Continuous ingestion across dozens of alternative and traditional streams, normalized into a longitudinal, point-in-time view of each company.

Traditional market & fundamental data
High-frequency alternative data streams
Unstructured public text at web scale
Corporate events & disclosures

AI Modeling Core

The heart of the company: a proprietary model stack that fuses heterogeneous raw data into company health grades and the insights behind them.

NLP models for unstructured text
ML-based entity resolution at scale
Cross-source signal fusion
Ensemble time-series forecasting

Insight Delivery

Grades and insights delivered the way analytical teams consume them — structured, versioned, and point-in-time correct.

Company health grades, updated weekly
Seven interpretable signal families
API & data-feed delivery
Full point-in-time history for research

Growth momentum Execution quality Demand signals Competitive positioning Operational efficiency Narrative & sentiment Risk & resilience

How it works

From millions of raw records to a clear company grade

Every week, the platform re-reads its full data universe end to end. Four stages, fully automated, cloud-native.

01 · INGEST

Collect

Tens of millions of data points per month across alternative and traditional sources — structured feeds, public documents, high-frequency streams — ingested and quality-scored on arrival.

02 · RESOLVE

Structure

ML-based entity resolution maps every record to the right company and point in time, reconciling inconsistent identifiers across heterogeneous sources into one clean longitudinal graph.

03 · UNDERSTAND

Interpret

Transformer-based NLP models read unstructured public text at scale, extracting themes, tone, and operational evidence that no keyword system can catch.

04 · GRADE

Grade

Ensemble models fuse all signal families into an overall health grade and per-dimension insights for each company — continuously validated against subsequent real-world outcomes.

AI at the core

AI isn't a feature of our product. It is the product.

Kipi Technologies exists because this problem cannot be solved without machine learning: the inputs are massive, unstructured, noisy, and fragmented. Every layer of the platform is a model.

Language models for unstructured data

The richest inputs we process are text. Transformer-based models perform aspect-level analysis, theme extraction, and tone scoring across millions of public documents — handling context, nuance, and domain-specific language that defeats classical NLP.

Entity resolution at scale

Alternative data is fragmented and inconsistent by nature. Our matching models resolve companies and events across heterogeneous sources into a single coherent graph — the foundation every downstream model depends on.

Grading ensembles with honest evaluation

Company grades come from ensembles of gradient-boosted and deep time-series models, trained on years of point-in-time data. Every model version is evaluated against strict walk-forward tests before it ships — no look-ahead, no survivorship bias.

Cloud-native training & inference

The full pipeline — ingestion, feature computation, model training, and weekly inference across the coverage universe — runs on managed cloud infrastructure, letting a small team retrain and re-grade hundreds of companies routinely.

Who it's for

Built for teams that need to understand how companies are really doing

Company research

Research teams use Kipi grades as an independent, continuously updated read on company health — a structured complement to the datasets everyone already has.

Competitive intelligence

Strategy teams track how the companies around them are really executing — momentum, efficiency, and positioning quantified consistently and comparable across a sector.

Market & sector research

Analysts and researchers use aggregated signal histories to study how momentum builds and fades across the market — and where it's building next.

Data & ethics

Sourced legitimately, handled responsibly

Trust in the data is trust in the signal. Our data practices are simple and strict.

Legitimate sources only

Every input is publicly available or properly licensed — public documents, disclosures, and compliant data streams. No scraping behind logins, no gray-market data, no privately obtained information.

Company-level outputs only

Signals describe companies, never people. Where source data originates with individuals, it is used solely as anonymous, aggregated input — no individual is identified, profiled, scored, or resold, ever.

Methodological integrity

Point-in-time data discipline, walk-forward evaluation, and versioned models. We hold our signals to the standard our clients hold their own research to.

Where we're going

The roadmap: alternative data as a standard layer of company analysis

Traditional financial data had its infrastructure revolution decades ago. Most of the data the world produces about companies still goes unread. We're building the models — and the market — that change that.

Live today

US coverage, weekly company grades

Hundreds of US companies graded weekly across seven signal families, with multi-year point-in-time history and production API delivery.

Broader coverage & inflection-event detection

Extending coverage across the broader US company universe, and adding real-time detection of operational inflection events — the moments when a company's health visibly changes, flagged as they happen.

Ahead

A conversational research layer

A generative-AI research interface over the signal base: ask any question about any company's health and direction in plain language, grounded in our models' evidence.

Team

Founders

Kipi Technologies was founded by a builder of large-scale data systems and a finance and operations veteran — the two disciplines this product demands.

Itay Keller

Co-founder

Veteran engineering leader (Google, Dell) with 13 patents in distributed systems. Architect of Kipi's signal-processing pipeline, data platform, and ML model stack.

LinkedIn keller@kipi-tech.com

Oded Segal