AI-Centric Philosophy

From Raw Data to
Decision-Ready Products

DataKnobs uses AI as a transformation layer to convert raw financial and market data into explainable, validated, decision-ready data products.

Executive Summary

At DataKnobs, we treat AI as the transformation layer that converts raw data into decision-ready data products.

Raw Data
AI Transformation Layer
Data Products (Signals → Decisions)

A successful data product is:

Algorithmically Valid

User-Adopted

Explainable with Lineage

Framework Architecture

1

Source Data Layer

Earnings calls
Financial metrics
Options data
Market data
Raw, unstructured, fragmented
2

AI Transformation Layer

The Core Engine

HEART OF THE SYSTEM

AI combines multiple capabilities to transform data:

A. Generative AI

  • Summarizes earnings calls
  • Extracts insights, risks, themes
  • Explains signals naturally
Example: "Revenue growth driven by pricing, but margin pressure emerging"

B. Agentic AI

  • Automates workflows
  • Fetches new data & triggers pipelines
  • Orchestrates computations
Example: Pull latest earnings → Recompute CPS → Update zones

C. ML Models

  • Predict probabilities
  • Score signals
  • Detect patterns
Example: Momentum prediction, CPS scoring, Breakout probability

D. Computation Layer

  • Business logic & Feature engineering
  • Signal combination
  • Scoring formulas
CPS = f(momentum, fundamentals, sentiment)
CTS = conviction × timing
Key Insight

The transformation layer is not one model—it is a system of AI + logic + pipelines working together.

3

Data Product Layer

Outputs

Atomic Products

  • AI Earnings Insight
  • Momentum Score
  • CPS
  • Options Demand/Supply

Composite Products

  • Conviction Signal
  • Timing Signal

Decision Product

  • Conviction-Timing Score (CTS)
  • BUY HOLD WAIT EXIT
NEW

4. Lineage & Explainability

"How was this produced?" Every product must answer this to build trust and debuggability.

Lineage Includes:
Source data Transformation steps Models applied Formulas used Intermediate signals
Example (CPS Lineage):
Earnings Data → Momentum → Sentiment → Formula → CPS Score
Example (CTS Lineage):
CPS + Momentum + Options Bias + Supply/Demand → CTS → Buy
NEW

5. Validation Layer

Dual validation ensures both model accuracy and real-world utility.

A. Signal Validation

  • • Backtesting
  • • Predictive accuracy
  • • Statistical validation

B. User Validation

  • • Workflow integration
  • • UI + AI interactions
  • • Adoption metrics
  • • Decision impact
NEW

6. User Interaction Layer

Data products are delivered seamlessly through native workflows:

Stock Webpages
Dashboards
Conversational AI
Example Interaction
User views CPS & Demand/Supply, asks:
"Why is CPS high? What changed last quarter?"
AI explains using Lineage + Insights:
"CPS increased because despite a slight revenue miss, forward momentum on cloud margins jumped 14% and options sentiment shifted bullish."

Key Differentiators

The DataKnobs philosophy stands apart from traditional data pipelines.

1. AI as Transformation

Not a UI gimmick. AI is deeply embedded in computation and pipeline orchestration.

2. Layered Data Products

A structured progression: Signals → Insights → Decisions.

3. Dual Validation

Rigorous validation of both mathematical model accuracy and actual user adoption.

4. Built-in Lineage

Complete transparency. Every output is 100% explainable to the user.

5. Workflow-Centric Design

Insights are delivered directly where decisions are made—via Web UI and Conversational AI, seamlessly integrating into daily routines.

Visual Architecture

Slide-ready representations of the DataKnobs Philosophy

AI Transformation Architecture

DATA PRODUCTS

Signals → Insights → Decisions (CTS)

AI TRANSFORMATION LAYER

GenAI
(Insights)
Agentic AI
(Automation)
ML Models
(Prediction)
+ Computation / Logic Layer
(Formulas, Pipelines, Feature Engg)

SOURCE DATA

Earnings | Financials | Options | Market

+ Lineage + Validation + User Feedback

End-to-End System

RAW DATA
AI TRANSFORMATION
GenAI | Agents | ML | Logic
DATA PRODUCTS
Signals → Insights → Decisions
USER INTERACTION
Web App | Dashboard | AI Q&A
VALIDATION LAYER
Model Accuracy + User Adoption
FEEDBACK LOOP

Key Concept (Executive)

DataKnobs Philosophy
Raw Data
AI Transformation Layer
Data Products
(Signals → Decisions)
Engine AI
Output Data Product

What makes this powerful?

You now have a complete, modern data product philosophy.

Just pipelines
But products
Just AI
But AI as a system
Just signals
But decision intelligence
Just models
But validated + adopted outputs
Just outputs
But explainable lineage