The Architecture Was PublishedBefore the Product Was Built.
Peer-reviewed research is not a credential. It is a moat. Competitors cannot claim prior art on a system they did not invent.
Designing Explainable Digital Decision Systems for Venture Evaluation and Public Funding
Kartik Kashyap · IJFMR Vol. 7, Issue 6 · Nov–Dec 2025 · ISSN: 2582-2160
- • Formal definition of the Evaluation Gap in Indian institutional capital markets
- • Architecture specification for explainable AI decision systems
- • The case for evidence-anchored scoring with deterministic arithmetic
- • Design-Science Research methodology across 7 prototype systems
Algorithmic Dissent: An Evidence-Anchored Multi-Agent Framework for Venture Evaluation
Kartik Kashyap · Target: Expert Systems with Applications (Elsevier) · 2026
The Master Formula: Vₛ = (wₘSₘ + wfSf + wₜSₜ) × (1 − Pf) × (1 − Gr)
TTR Metric: Time-to-Replicate — quantifying technical moat depth in months with explicit funding assumptions.
Fixed Published Weights: Market 40%, Financial 35%, Technical 25%. The weighted base is multiplied by (1 − failure risk) and (1 − geopolitical risk).
Evidence Boundary: Missing company data is flagged rather than replaced with industry averages. Model-produced inputs can vary, so the number is not a validated measure of startup quality.
The Seven Sensors —The Research History
Before Zurvek, VentureSense developed and killed seven prototype systems. Each generated specific findings about where institutional evaluation fails.
| System | Domain | Key Finding |
|---|---|---|
| InsightVault | Market trend intelligence | Trend data without evaluation framework produces noise, not signal |
| Authntk8 | Biometric trust verification | Identity verification is necessary but insufficient for investment decisions |
| Chetara | Temporal trend analysis | Recency bias in trend data systematically distorts market sizing |
| Vencrypt | Asset scoring | Scoring without adversarial challenge produces optimistic bias |
| NIRNAE | Government audit intelligence | Public sector evaluation requires sovereign data integration unavailable to generic AI systems |
| LIKHIT | Internal operations | Operational intelligence is a second-order problem; evaluation is first-order |
| FORESY8 | News awareness | News signal without forensic framework produces narrative, not judgment |
Advisors
Dr. Dinesh Jain
Associate Professor, IIPMB Bengaluru
Public Policy, MSME Funding Systems, Institutional Finance
Dr. Haribabu Venketeswaran
Joint Director, Andhra Pradesh Innovation Society
ML Pipelines, Innovation Ecosystem, AI Strategy
Guides
Checklist
VC Due Diligence Checklist — 99-point framework
The seven institutional pillars investors use to evaluate startups.
Guide
How VCs Use AI for Due Diligence
Adversarial evaluation, claim extraction, and the committee brief.
Guide
The Venture Capital Investment Process: End-to-End Guide
Sourcing, screening, diligence, IC, term sheet, close, and post-investment.
Guide
Venture Capital Software: A Practical Guide for Modern Funds
The tooling stack behind sourcing, diligence, portfolio, and LP reporting.
Comparison
Best AI Due Diligence Software (2026)
Comparison guide for VC and institutional investors.
Chronology
The Long Arc of Financial Innovation
From Lydian coins to unified ledgers — 26 centuries in one read.
Analysis
The Bandwidth Problem: India's Accelerators vs Their Deal Flow
1,100+ incubators, thousands of applications, and the screening capacity that never arrived.
Teardown
Investment Memo Teardown: What Each Section Is For
The six working parts of an IC memo, the evidence under each, and what it does not claim.
India
India Startup Due Diligence: The Evidence Layer
Filings, instruments, GST reconciliation, statutory dues, licensing and reference practice.
Process
How Much Diligence Is Enough? Depth by Cheque Size
Proportionate work at angel, seed, Series A and growth — and what is theatre.
Recognition
Inquiries: support@zurvek.com · VentureSense Technologies LLP · LLPIN: ACR-5230