PRIVATE CAPITAL
Deal Origination
Diligence & Value Creation
ESG/Climate Risk
Private Capital Intelligence Platform that couples GenAI scouting & document copilots with Causal ML & uplift experimentation for operational playbooks, and portfolio climate/ESG analytics aligned to PCAF/SBTi standards.
GenAI + Causal ML
RL + Open Data
Repeatable Edge
Proprietary Deal Flow
Defensible Analytics
Auditable Components
Private Capital Intelligence
Deal Origination
2-4×
vs banker funnels
Diligence Speed
-70%
time to red-flag memo
Pricing ROI
+8pp
causal margin uplift
ESG Coverage
≥90%
PCAF AUM coverage
GenAI Scouting
Active
Causal ML Engine
Learning
Climate Risk
Monitoring
Four Pressure Points
Private equity, growth equity, and credit investors need a repeatable edge
01
Proprietary Deal Flow at Scale
Surfacing sub-(10)K "hidden" targets from fragmented signals
Challenge: Not just banker pipelines
Filings, job posts, product docs, reviews - fragmented signals across multiple sources
02
Faster, Deeper Diligence
Compressing weeks of red-flag hunting into days
Challenge: Without missing "known unknowns"
10-Ks/8-Ks, contracts, and tech repos analysis with comprehensive coverage
03
Proof of Value Creation
Quantifying what works after close with causal estimates
Challenge: Not correlations - withstands IC scrutiny
Pricing, product, go-to-market, operations with causal proof points
04
Credible ESG & Climate Risk
Portfolio-level financed-emissions accounting and risk views
Challenge: Aligned to current standards and investor expectations
Transition/physical-risk views aligned to PCAF/SBTi standards
Tensorblue's Solution
A Private Capital Intelligence Platform that couples (A) GenAI scouting & document copilots with (B) Causal ML & uplift experimentation for operational playbooks, and (C) portfolio climate/ESG analytics aligned to PCAF/SBTi plus (optional) catastrophe-loss modeling via Oasis LMF. Everything runs on open, auditable components investors can diligence.
Real, Open Anchors We Build On
Auditable today - proven foundations for private capital intelligence
01
Financial LLM
FinGPT
Open financial LLM family with automated data curation and lightweight adaptation; designed to keep models fresh against internet-scale financial data.
Source: GitHub - AI4Finance-Foundation
Use Case: Parse filings, calls, and news streams in scouting/diligence copilots
Open & Auditable
02
Analytics Stack
OpenBB Platform
Open analytics stack and connectors (equity, macro, credit, FX, alternative data) for target screens, comparables, and factor tilts.
Source: GitHub - OpenBB-finance
Use Case: Python and API interfaces for multi-asset coverage
Open & Auditable
03
Regulatory Data
SEC EDGAR APIs
Authoritative APIs for submissions & extracted XBRL to power filing parsers and anomaly detectors in IC memos.
Source: SEC.gov
Use Case: Filing parsers and anomaly detection for diligence
Open & Auditable
04
Causal Inference
Causal ML Toolchain
EconML (Microsoft), DoWhy (PyWhy), DoubleML (Chernozhukov et al.) for heterogeneous treatment effects, policy learning, and debiased estimates.
Source: Microsoft Research
Use Case: Treatment effects that survive robustness checks
Open & Auditable
05
Catastrophe Modeling
Oasis Loss Modelling Framework
Open catastrophe modeling stack to assess physical-risk loss distributions for portfolio assets (e.g., coastal DCs, plants, warehouses).
Source: oasislmf.org
Use Case: Physical-risk assessment for portfolio assets
Open & Auditable
06
ESG Standard
PCAF Standard
Current, widely used methodology to quantify financed emissions across asset classes; we implement its disclosure checklist for portfolio reporting.
Source: carbonaccountingfinancials.com
Use Case: Financed emissions quantification and reporting
Open & Auditable
07
Climate Framework
SBTi / VCMI Context
Evolving guidance for financial institutions and Scope-3/portfolio decarbonization (SBTi tightening for FIs; VCMI's 2025 Scope-3 Action Code).
Source: Reuters
Use Case: Portfolio decarbonization alignment and tracking
Open & Auditable
Credible Foundation
These open, auditable components provide transparent, inspectable foundations that investors can diligence, ensuring credibility and reproducibilityacross all private capital intelligence operations.
End-to-End Architecture
Private capital intelligence data flow
Signals Layer
EDGAR/XBRL
SEC filings
OpenBB Feeds
Market data
Job Posts
Hiring signals
Web Reviews
Sentiment data
Patents
IP intelligence
Repos
Tech stack
News
Media coverage
↓
Data Fabric
Ingestion (APIs, web), entity resolution, cleaning, embeddings
Vector store (targets, docs, calls)
Feature store (rev, margin, cohort)
Risk data (hazard, exposure, emissions)
↓
Causal & Optimization Layer
EconML/DoWhy/DoubleML
Uplift & ATE/HTE
RL/BO
Pricing & media mix
Oasis LMF
Catastrophe loss sims
PCAF Calculator
SBTi/VCMI alignment
↓
Applications & IC Interfaces
Deal Scout
Web
Diligence Assistant
Data room
Value Creation Workbench
A/B, quasi-experiments
Climate/ESG Cockpit
Portfolio & asset
Mobile Portfolio Monitor
Mobile
Pillar A: GenAI Scouting & Diligence Copilots
Fresh, grounded intelligence for deal origination and due diligence
1
Target Surfacing
LLM-guided retrieval over OpenBB + web/job/patent signals; filters on growth, margins, hiring momentum, tech stack, and tell-tale disclosure phrases.
"OpenBB connectors give immediate multi-asset coverage"
Growth metrics
Margin analysis
Hiring momentum
Tech stack detection
Disclosure analysis
2
Doc Copilot (EDGAR-native)
Parses 10-K/10-Q/8-K and exhibits with FinGPT adapters; flags changes in revenue recognition, segment reclass, unusual KPIs.
"Links to XBRL facts and footnotes via EDGAR APIs"
10-K/10-Q parsing
8-K analysis
Revenue recognition
Segment reclassification
KPI anomaly detection
3
Tech Diligence
Optional GitHub/issue tracking scans (for software targets) and SBOM-style summaries.
"Software bill of materials and tech stack analysis"
GitHub analysis
Issue tracking
SBOM generation
Tech stack mapping
Dependency analysis
4
Signals View
Unknowns map (what we still don't know), competitive mentions, founder churn, procurement footprints.
"Comprehensive intelligence gaps and competitive analysis"
Unknowns mapping
Competitive mentions
Founder churn tracking
Procurement analysis
Intelligence gaps
Why It's Credible
FinGPT designed for finance data freshness
EDGAR APIs ensure traceable citations
OpenBB provides reproducible data pulls
Your team can inspect all data sources
No black box - fully auditable components
Data Sources
EDGAR Filings
10-K, 10-Q, 8-K, proxy statements
OpenBB Market Data
Equity, macro, credit, FX data
Alternative Signals
Jobs, patents, reviews, news
Tech Intelligence
GitHub, tech stack, dependencies
AI Components
FinGPT Adapters
Financial domain-specific language models
RAG Pipeline
Retrieval-augmented generation for accuracy
Vector Embeddings
Semantic search across documents
Citation Tracking
Every claim backed by source documents
Pillar B: Causal Value-Creation Engine
Prove ROI before & after close with causal evidence
The Problem
Portfolio teams often ship dashboards of correlations; ICs and LPs want causal proof.
"This is where most "AI in PE" pitches fail; we pass because we use the causal stack above."
1
Pricing & Promo
METHODS
EconML metalearners + DoubleML for debiasing
Description
Run uplift modeling on past offers; deploy policy learners to recommend personalized prices/discount ladders; run prospective AB tests where feasible.
Outputs
Treatment policies with confidence intervals
Bias diagnostics
Don't treat segments
IC-ready notebooks
2
Sales Acceleration
METHODS
DoWhy DAGs formalize assumptions you can defend in IC memos
Description
Heterogeneous treatment effects by segment/channel to decide where SDR headcount moves the needle.
Outputs
Segment-specific treatment effects
Channel optimization
Resource allocation guidance
Causal DAG documentation
3
Digital Operations
METHODS
Synthetic control methods + staggered difference-in-differences
Description
Causal impact of new logistics SLA or warehouse slotting algorithm; staggered adoption designs + synthetic controls.
Outputs
Operations impact estimates
Implementation roadmaps
ROI projections
Risk assessments
4
Media Mix / CAC
METHODS
Bayesian optimization + Reinforcement learning
Description
Bayesian optimization with guardrails; RL where feedback is rapid (e-comm).
Outputs
Media mix optimization
CAC reduction strategies
Budget allocation
Performance predictions
Causal ML Technical Stack
EconML
Microsoft Research
DoWhy
PyWhy Foundation
DoubleML
Chernozhukov et al.
Bayesian Optimization
GPyTorch + BoTorch
Defensible Results
Confidence Intervals
Statistical significance for all treatment effects
Bias Diagnostics
Robustness checks and sensitivity analysis
Version Control
IC-ready notebooks with full reproducibility
Assumption Testing
DoWhy DAGs formalize defendable assumptions
Business Impact
Pricing Optimization
+3-8pp margin uplift with causal proof
Media Mix Efficiency
-10-25% CAC reduction with confidence bounds
Resource Allocation
Data-driven headcount and budget decisions
LP Reporting
Causal evidence for value creation claims
Pillar C: Portfolio Climate & Catastrophe Risk
ESG that matters - aligned to current standards and investor expectations
1
Financed Emissions (PCAF)
"Portfolio-level carbon accounting aligned to global standards"
Description
Compute financed emissions by asset class; attach a PCAF disclosure checklist to each fund/vehicle; reconcile to LP reporting packs.
Key Features
Asset class emissions
PCAF disclosure checklist
LP reporting reconciliation
Automated calculations
2
Targets (SBTi) & Offsets Policy (VCMI)
"Science-based targets and voluntary carbon market integration"
Description
Portfolio-level progress vs SBTi for FIs; where residual Scope-3 exists, show VCMI-aligned use of high-quality credits with caps/guardrails.
Key Features
SBTi alignment tracking
VCMI compliance
High-quality credits
Residual scope-3 management
3
Physical Risk via Oasis LMF
"Catastrophe modeling for portfolio asset risk assessment"
Description
Run cat-model views (wind, flood, quake) for plants/DCs; link exceedance-probability loss curves to debt covenants and insurance terms.
Key Features
Wind/flood/earthquake modeling
Exceedance probability curves
Debt covenant analysis
Insurance optimization
Standards & Frameworks Alignment
PCAF
Partnership for Carbon Accounting Financials
Carbon Accounting
SBTi
Science-Based Targets initiative
Target Setting
VCMI
Voluntary Carbon Markets Integrity Initiative
Carbon Markets
TCFD
Task Force on Climate-related Financial Disclosures
Risk Disclosure
Climate Risk Coverage
Physical Risk Assessment
Transition Risk Analysis
Carbon Footprint Tracking
Climate Scenario Modeling
Insurance Optimization
Regulatory Compliance
Carbon Accounting Engine
PCAF Implementation
Global standard for financed emissions calculation
Automated Disclosures
LP-ready reporting with audit trails
Scope 3 Tracking
Portfolio-wide indirect emissions monitoring
Benchmarking
Industry comparisons and peer analysis
Physical Risk Modeling
Oasis LMF Integration
Open-source catastrophe modeling framework
Multi-Hazard Analysis
Wind, flood, earthquake, wildfire modeling
Loss Curves
Exceedance probability and return period analysis
Insurance Optimization
Coverage analysis and premium optimization
Expected ESG Impact
≥90%
PCAF Coverage
Of AUM with automated disclosures
100%
Priority Sites
Modeled with Oasis LMF
Real-time
Compliance Monitoring
SBTi/VCMI framework tracking
Web & Mobile Apps
Partner-ready UX for private capital operations
01
Deal Scout (Web)
Description
Pipeline heatmap by thesis; one-click "why surfaced" with citations (filings §§, job stratification, alt-signals).
Pipeline heatmaps
Thesis-based filtering
Citation tracking
Alternative signals integration
Example
"One-click explanation of why each target was surfaced with full audit trail"
02
Diligence Data-Room Assistant
Description
Ask: "Show revenue recognition changes from 2022→2024 and cross-reference ASC 606 risks." — returns excerpts + XBRL anchors.
Natural language queries
Document cross-referencing
XBRL anchor links
Risk identification
Example
"EDGAR-native analysis with traceable citations to filing sections"
03
Value-Creation Workbench
Description
Pick a portfolio company → select playbook (pricing, churn, supply chain) → see causal uplift estimates, suggested policy, and risk/uncertainty bounds.
Portfolio company selection
Playbook library
Causal uplift estimates
Risk/uncertainty bounds
Example
"Causal treatment effects with confidence intervals for IC memos"
04
Climate/ESG Cockpit
Description
Financed-emissions waterfalls (PCAF), SBTi/VCMI status, Oasis EP curves per site for lender/insurer conversations.
Financed emissions waterfalls
SBTi/VCMI status tracking
Oasis exceedance curves
Lender/insurer reporting
Example
"Portfolio-level climate risk dashboard for LP reporting"
05
Mobile (Partners/Ops)
Description
Portfolio KPIs, alerts on filing anomalies, covenant-risk pings (e.g., flood EP@1% updated).
Portfolio KPIs
Filing anomaly alerts
Covenant risk monitoring
Real-time notifications
Example
"Mobile alerts for critical portfolio events and risk updates"
Platform Features
Audit-Ready
Every decision backed by traceable data sources and citations
Mobile-First
Optimized for partners and operations teams on-the-go
IC-Ready
Investment committee presentations with causal evidence
User Experience Highlights
Investment Teams
• Natural language queries for complex diligence questions
• Causal treatment effects with confidence intervals
• One-click explanations with full citation trails
• Portfolio-level climate risk dashboards
Operations & Partners
• Mobile alerts for critical portfolio events
• Real-time KPI monitoring and reporting
• Covenant risk notifications and updates
• Filing anomaly detection and alerts
Data, Infrastructure & Governance
Industrial-grade infrastructure for private capital intelligence
| Layer | Tools / Notes | Purpose |
|---|---|---|
Ingestion | OpenBB connectors; SEC EDGAR APIs; alt-signals (jobs, patents, reviews) | Multi-source data collection |
Storage | Delta Lake; document store + vector DB for RAG | Time-series and document storage |
Feature store | Feast (cohorts, pricing, CAC, unit economics) | ML feature management |
Modeling | FinGPT adapters; EconML/DoWhy/DoubleML for causal; RL/BO for pricing/media | AI/ML model development |
Risk | Oasis LMF for cat losses; PCAF calc engine; SBTi/VCMI rules | Climate and risk modeling |
Serving | Triton/FastAPI; notebook exports for IC | Production deployment |
Security | Role-segmented workspaces per deal; full citation/audit trails | Access control and compliance |
Multi-Source Ingestion
OpenBB connectors, SEC EDGAR APIs, and alternative signals (jobs, patents, reviews) provide comprehensive data coverage for deal origination and diligence.
Auditable Components
All AI/ML models built on open, inspectable components (FinGPT, EconML, DoWhy, DoubleML) that investors can diligence and verify.
Security & Compliance
Role-segmented workspaces per deal, full citation/audit trails, and compliance with financial data handling requirements.
Governance Framework
Data Governance
• Role-based access control per deal/portfolio
• Full audit trails for all data access and model usage
• Citation tracking for all AI-generated insights
• Version control for all models and analyses
Compliance & Security
• Strict scope to public + client-provided data
• MNPI protection and leakage prevention
• Regulatory compliance (SEC, financial regulations)
• ESG framework alignment and tracking
KPIs You Can Defend
Measurable improvements backed by proven toolchains
| Area | KPI | Typical Improvement Band | Reference |
|---|---|---|---|
Deal Origination | Qualified proprietary targets / month | 2–4× vs banker-only funnels | FinGPT+OpenBB+signals |
Diligence Speed | Time to red-flag memo | −50–70% | EDGAR-grounded copilot |
Pricing/Media ROI | Causal uplift on margin / CAC | +3–8 pp margin; −10–25% CAC | EconML/DoubleML playbooks |
Climate Reporting | PCAF coverage of AUM | → ≥90% | Financed-emissions coverage with audit pack |
Physical Risk | Sites with EP curves / insurance leverage | 100% | Priority sites modeled (Oasis) |
2-4×
Deal Flow Multiplier
Proprietary target identification
-70%
Diligence Speed
Time to red-flag memo
+8pp
Margin Uplift
Causal pricing optimization
Why These KPIs Are Defensible
Open Source
Built on auditable, open components
Academic Rigor
Peer-reviewed causal ML methods
Regulatory Data
SEC EDGAR and authoritative sources
Reproducible
Full methodology transparency
* Bands reflect what the cited toolchains enable; actuals vary by data quality and adoption.
Case Snapshots (Anonymized)
Patterned on real deployments across private capital strategies
01
Buy-and-Build (Industrial Tech)
Scouting copilot surfaced 37 off-market bolt-ons in 6 weeks
Off-Market Targets
37
Screening Pass Rate
11 passed
LOIs Generated
3
Gross Margin Lift
+420 bps
Scouting copilot surfaced 37 off-market bolt-ons in 6 weeks; 11 passed screen; 3 LOIs. Diligence assistant flagged an 8-K rev-rec change that altered earn-out. Post-close pricing policy (DoubleML) lifted gross margin +420 bps.
02
Consumer Growth Equity
Media-mix RL trimmed blended CAC while preserving LTV
CAC Reduction
−18%
LTV Preservation
Maintained
No-Treat Segments
2 identified
Quarterly Savings
$1.2M
Media-mix RL trimmed blended CAC −18% while preserving LTV; EconML showed no-treat for two segments, saving $1.2M/quarter in wasted spend.
03
Infrastructure Roll-Up
Oasis cat risk revealed 1% AEP losses above covenants
AEP Loss Analysis
1% above covenants
DCs Analyzed
2 sites
Insurance Terms
Improved
PCAF Coverage
96% of AUM
Oasis cat risk revealed 1% AEP losses above covenants for two DCs; insurance negotiation improved terms; PCAF coverage reached 96% of AUM with automated disclosures.
Implementation Pattern
All cases demonstrate GenAI-powered deal originationwith auditable citations, causal value creationwith statistical rigor, and ESG/climate riskaligned to current standards — all built on open, inspectable components that investors can diligence.
Risks & Mitigations
Risk
GenAI hallucination in diligence
Mitigation
RAG over EDGAR; every claim shows filing section/XBRL anchor; human review.
Risk
Causal wrong-sign estimates
Mitigation
DAG design in DoWhy; placebo & sensitivity checks; DoubleML orthogonalization; report CIs.
Risk
Data leakage / MNPI
Mitigation
Strict scope to public + client-provided; audit trails; red-team prompts for leakage.
Risk
ESG framework drift
Mitigation
Track SBTi/VCMI updates; versioned rules; change-logs in cockpit.
Implementation Roadmap
8–12 weeks to first wins
1
Weeks 0–2
Foundation Setup
Description
Connect OpenBB + EDGAR; stand up vector store; baseline scout screens.
Key Deliverables
OpenBB connectors configured
SEC EDGAR API integration
Vector database setup
Initial scout screening models
2
Weeks 2–5
Diligence Copilot
Description
Diligence copilot with FinGPT adapters; first red-flag memos.
Key Deliverables
FinGPT adapters deployed
Document parsing pipeline
Red-flag detection system
First IC memos generated
3
Weeks 4–8
Causal Workbench
Description
Causal workbench on 1–2 portcos (pricing/churn); ship policy with guardrails.
Key Deliverables
EconML/DoWhy integration
Portfolio company models
Treatment policy recommendations
Confidence interval reporting
4
Weeks 6–10
Climate & ESG
Description
PCAF engine + Oasis cat runs for top 10 assets; LP-ready dashboards.
Key Deliverables
PCAF calculation engine
Oasis LMF integration
Climate risk dashboards
LP reporting automation
5
Weeks 10–12
Mobile & Optimization
Description
Mobile portfolio monitor; A/B experimentation cadence; playbook library.
Key Deliverables
Mobile portfolio app
A/B testing framework
Playbook library
Full deployment
Success Metrics by Phase
Week 2
Foundation
First proprietary targets surfaced
Week 5
Diligence
First red-flag memo generated
Week 8
Value Creation
First causal treatment policy
Week 10
ESG
First LP climate report
Week 12
Full Platform
Complete system operational
Implementation Approach
Each phase builds on the previous with incremental value delivery. Early phases focus on proven open-source components(OpenBB, FinGPT, EconML) for rapid deployment and immediate wins, while later phases integrate more complex causal and ESG capabilities.
Defensible analytics spine for GPs
Tensorblue gives GPs a defensible analytics spine — fresh, sourced deal intelligence; diligence that cites filings; value-creation decisions with causal proof; and ESG/climate metrics aligned to the standards LPs ask about today(PCAF, SBTi/VCMI) — all on open, inspectable tech.
Key Differentiators
Fresh Deal Intelligence
2-4× proprietary targets vs banker funnels
Causal Proof
Value creation with statistical rigor
Auditable Citations
Every claim backed by SEC filings
ESG Standards
PCAF/SBTi/VCMI aligned reporting
Result: repeatable edge — combining proprietary deal flow, causal value creation, auditable diligence, and ESG compliance for measurable competitive advantage