AI where it counts.
Rigor everywhere else.
The Koi Engine builds avoided emissions models on demand, turning solution descriptions into customized, auditable forecasts backed by years of methodology development and a curated climate data lake. It produces in seconds what used to take weeks of manual analysis.
How It Works
Four stages. Full transparency at every one.
The engine does not replace scientific judgment. It accelerates the structured, repeatable parts so scientists can focus on the decisions that actually require expertise.
Identify
Maps each climate solution to the baseline it displaces and the value chains it touches. No forecast is credible without knowing what you are comparing against.
Draws from curated datasets spanning IEA, SBTi, US Federal LCA Commons, and peer-reviewed literature to select the right reference points.
Match
Aligns identified value chains to GHG intensity and market data in the Koi Data Lake. Enforces dimensional analysis so intensity multiplied by market size produces meaningful avoided emissions.
Automatic unit conversion across energy (Wh, J, BTU), mass, distance, and economic value. Incompatible units are caught, not papered over.
Fill
Where the Data Lake lacks a specific data point, AI synthesizes relevant literature to fill the gap. Every AI-sourced value is flagged on the model Datasheet.
Market capture estimation uses SEC filings and sales data to approximate real deployment timelines. This is the first automated method that avoids blanket market uptake assumptions.
Produce
Outputs a fully customizable, version-controlled model with transparent inputs, documented assumptions, and audit-grade provenance.
Every model is evergreen. When underlying Data Lake information changes, models update automatically, keeping forecasts consistent across technologies and time.
Grounded in established data
Scaffolding & Guardrails
AI accelerates. Methodology governs. The engine is where they meet.
Close to a decade of climate modeling expertise defines every constraint, every validation, and every default the engine operates within. AI accelerates the execution. It does not make the decisions.
Pre-Validated Data Lake
Curated sources (IEA, SBTi, US Federal LCA Commons, academic literature) form the foundation. The engine reasons on top of verified data, not raw web scrapes.
Unit Compatibility Enforcement
Dimensional analysis is enforced at every stage. If emissions per mass meets a market denominated in value, the engine resolves or rejects. It never silently proceeds.
AI Data Flagging
Every value sourced by AI is explicitly marked on model Datasheets. Scientists and users know exactly which data points were synthesized and which came from the Lake.
Plausibility Bounding
Automated models are bounded for plausibility before release. Three QA tiers (Rapid, Refined, Diligence) layer increasing human review on top of engine output.
Auditability
Every input traceable. Every assumption documented.
Speed means nothing if you cannot defend the output. The engine is built so that every forecast can withstand the scrutiny of investors, auditors, and regulators.
Version Control
Every model versioned. Select prior or future versions with full change tracking.
Transparent References
Data sources, boundaries, and assumptions documented. Nothing is hidden behind a black box.
Edit Mode
Review, modify, and customize any model while maintaining the audit trail. Your changes, your reasons, all on the record.
QA Datasheets
Tags indicate manual scientist review level, data verification status, and AI-sourced values. Everything visible at a glance.
Viewing Version
1.1 LatestAccess Level
Owner
QA Tier
RefinedLast Updated
2026-01-21
Forecast Configuration
Annual results between 2026 and 2036
Addressable Market
milk production (1,100 Mt milk)
Market Capture
1% in 10 years (Start: 2025)
Baseline
Milk
Solution
Fermentation protein-based dairy alternative (thin consistency)
References
Current configuration references
Illustrative datasheet. Actual models contain full reference chains and methodology notes.
Years of methodology.
Seconds to a forecast.
See the engine in action on a solution you care about. Every input, every assumption, every output. No hand-waving.