Meridian

Unified Decision & Underwriting System

Institutional Look-Through & Bayesian Analytics for Mixed Public/Private Portfolios

"Meridian bridges the gap between sparse private-asset data and real-time public markets by combining production-grade software engineering with Bayesian risk analytics."

Unified Look-Through Single-pane risk visibility REITs · Direct CRE · Private Debt Bayesian Risk Modeling Probabilistic valuations Distributions, not point marks Production Engineering Sub-second analytics DuckDB + FastAPI in RAM
Scroll or use arrows

The Problem

The Allocator's Dilemma

Structural risk trapped in private assets — data sparsity, appraisal lag, and fragmented analytics leave allocators blind to true portfolio risk.

Public Markets gap ! Private Assets gap ! PDF Memos

  1. The Appraisal Smoothing Illusion

    Quarterly appraisals lag public markets by 3–6 months, artificially dampening reported volatility and inflating Sharpe ratios — masking real systemic tail risk.

  2. Unstructured Data Silos

    Critical covenant terms, debt maturities, and tenant escalation clauses remain trapped in 50+ page PDF deal memos and GP reports.

  3. Fragmented Analytics

    Liquid holdings tracked in real-time portfolio software while private deals live in isolated analyst DCF workbooks — leaving allocators blind to true correlations.

The Solution

In-Memory Peer-Compute Engine

Meridian ingests, indexes, and computes across all data formats — Excel DCFs, PDF memos, and live public feeds — in a single unified engine.

Multi-Modal Inputs

  • Analyst Excel DCFs
  • PDF Memos & Leases
  • Live Public Feeds

In-Memory Compute

  • FastAPI + DuckDB Core
  • PyMC / JAX Bayesian Math
  • Vectorized SQL Execution

Institutional Outputs

  • Factor Look-Through
  • Valuation Distributions
  • Cell-to-PDF Lineage

Non-Invasive

Analysts retain Excel. Meridian ingests via background APIs — no workflow disruption.

Sub-Second

DuckDB executes vectorized SQL in RAM for institutional-grade response times.

Full Lineage

Every metric traces to its exact cell or PDF page — cryptographic audit trail.

Quantitative Engine

Bayesian Un-Smoothing

Instead of treating quarterly appraisals as static ground truth, Meridian treats them as prior probability distributions updated continuously against live public market signals.

P(Vreal | Mpublic)  ∝  P(Mpublic | Vreal) · P(Vappraisal)
Posterior valuation given market signals  =  likelihood × prior appraisal distribution
PRIOR P(V_appraisal) Wide, uncertain update SIGNAL M_public Live market data yields POSTERIOR P(V_real | M_public) Narrow, calibrated

Appraisal Un-Smoothing

Recovers the true underlying return series, volatility, and cross-asset correlations hidden by stale marks.

Continuous Valuation Bands

Replaces single-point marks with dynamic 90% confidence intervals (e.g. $46.2M – $50.1M).

Regime Switching

Captures non-linear risk shifts during market disruptions using Markov-Switching models.

Decision Intelligence

CIO & Allocator Dashboard

Operationalizing look-through intelligence across a $1.2B mixed portfolio — revealing what traditional marks obscure.

ECON. NAV $1,184M Range: $1.15B – $1.21B vs. stated $1,210M RE EXPOSURE 34.2% 21% Direct + 13% REITs vs. stated 21% direct only SHARPE RATIO 1.12 vs. stated 1.48 (smoothed) True volatility realigned SHORT-TERM SHOCK 2.8× vs. 1.5× stated Debt maturity stress revealed

Factor Look-Through

Decompose public equities, private funds, and co-investments into fundamental macro factors.

Lineage Inspector

Click any valuation cell to highlight the exact clause in the GP quarterly report or Excel formula.

Scenario Stress Engine

Shock rates, cap rates, or occupancy — see instantaneous updates across liquid and illiquid holdings.

Why Meridian

Institutional Advantage Matrix

Meridian occupies the only position that unifies private-asset look-through, source traceability, and real-time public/private analytics in a single architecture.

Capability Legacy Systems AI / LLM Wrappers Meridian
Private Asset Look-Through ✕ Stale / Manual ✕ Hallucination-prone ✓ Bayesian Distributions
Public + Private Unification △ Siloed Modules ✕ Text-Only ✓ Real-Time Factor Model
Source Traceability ✕ None △ Unreliable ✓ Cryptographic Lineage
Data Privacy △ Vendor Lock-in ✕ API Exposure ✓ Zero-Durable-Plaintext

Zero-Durable-Plaintext Architecture

All indexing, vector search, and Bayesian sampling execute in RAM. Plaintext purged on session end. Runs within your private cloud or on-premise — behind your existing firewalls.

Deployment Options

Custom Build

12-week sprint. Full integration into your data environment.

Platform License

Core engine with custom plugin modules for your algorithms.

Advisory Retainer

Ongoing Bayesian modeling and quarterly IC stress-testing.