Skip to main content
Felixfusion
PILOT-READY · MULTI-AGENT RUNTIME

Quantitative decision intelligence for energy infrastructure.

Felixfusion is a multi-agent platform for high-stakes capital and regulatory work in energy. Two quantitative decision engines on the same runtime. Athena for energy: valuation and revenue intelligence on utility-scale BESS, hybrid renewables, and the data-center power deals reshaping load growth. Nexus for supply chain: sourcing, compliance, and contract intelligence across the equipment and EPC layer.

WHAT THE RUNTIME COVERS

The platform, by the numbers.

PLATFORM SNAPSHOT · v0.4STABLE COVERAGE METRICS · NOT MARKET DATA
Quantitative decision engines · Athena · Nexus · Energy and supply chain
2
Markets in runtime · ISOs and regions · NEM, ERCOT, GCC, PJM, roadmap
5
Use cases supported · decision shapes · 12 Athena · 8 Nexus
20
Buyer profiles served · decision shapes · 4 Athena · 4 Nexus
8
Decision stages · human-gated · Parse → State → Score → Draft → Approve
5
Pilot turnaround · weeks to first artifact · Single-engagement pilot
4–6
00 / ROUTING

Three routes through Felixfusion.

The engine powers two products serving distinct buyers, plus a diligence path for investors and technical counterparts. Pick the route that matches why you are here.

QUANTITATIVE DECISION ENGINE · ENERGY

Athena

Valuation and revenue intelligence for utility-scale BESS, hybrid renewables, and the data-center power deals reshaping load growth. Used by infrastructure PE, sovereign desks, IPPs, and hyperscaler energy teams.

QUANTITATIVE DECISION ENGINE · SUPPLY CHAIN

Nexus

Sourcing, compliance, and contract intelligence across BESS, solar, wind, and balance-of-plant equipment. Used by procurement leads, EPC originators, lender diligence groups, and project finance teams.

FOR INVESTORS AND TECHNICAL DILIGENCE

Architecture and Company

How the engine is built, what we ship, who we serve, and how we engage. Read these before pilot or diligence conversations.

01 / ENGINE - Built for the work nobody wants to do.

The engine is calibrated for the slowest, most paper-heavy work in infrastructure: a bankable IC memo on a co-located BESS, a tender response on a multi-billion-dollar federal procurement, an interconnection diligence pack on a renewable feeding a hyperscaler. Each of those engagements lives in document volume measured in thousands of pages, in clocks measured in weeks, and in counterparties that will sign their name to the output. The features below describe how the runtime survives that.

The engine survives the work nobody wants to do. Three operating constraints define the runtime — every feature below answers one of them.

  • Reads what humans won't. Grid cluster studies. Procurement mandates. Interconnection dockets. We parse the thousand-page filings that normally stall projects for months, then hand the relevant parts to your agents as structured context.
  • Holds state for weeks, not seconds. Real projects run on a quarterly clock, not a chat turn. Our agents keep context across handoffs, resume cleanly after interruptions, and leave an audit trail of every decision they made along the way.
  • A human always signs the final page. Nothing leaves the system without review. Every DD pack, every bid, every outbound artifact passes through a manual approval gate before it ever reaches a counterparty.
ENGINEERING SIGNALS

Five design pillars under the runtime.

The engine survives the work because of the design choices below, not because of model size. The architecture paper covers each in depth.

Read the architecture paper
Multi-agent runtime · weeks of state, not chat turns
Long-running
Data-agnostic ingestion · every claim links back to a page
Provenanced
Persistent task graph · checkpointed across restarts
Stateful
Approval gate · nothing leaves without a senior signature
Human-gated
Deployment posture · no shared multi-tenant data path
Customer-VPC
PIPELINE

From filings to signed memo.

Five stages, each gated by a human review before the artifact moves on. The pipeline runs identically for Athena IC memos and Nexus sourcing briefs.

  1. 01

    Filings

    Grid cluster studies, IC memos, regulatory dockets, asset performance reports, RFP packs.

  2. 02

    Parse

    Structured extraction across thousands of pages. Document provenance preserved.

  3. 03

    Score

    Lattice and Reservoir produce sensitivity bands, valuations, and risk ranks.

  4. 04

    Memo

    Bankable IC memo or compliance brief drafted against the parsed inputs.

  5. 05

    Sign

    Senior-analyst manual approval before the artifact reaches a counterparty.

02 / PRODUCT SUITE - Two quantitative decision engines on the platform.

The Felixfusion runtime powers both surfaces. Athena is the quantitative decision engine for energy. Nexus is the quantitative decision engine for supply chain. Both are built for teams where getting the answer wrong costs real money.

PRODUCT.01 // QUANTITATIVE DECISION ENGINE · ENERGY

Athena

The quantitative decision engine for energy. IC memos and revenue intelligence on utility-scale BESS, hybrid renewables, and the data-center power deals reshaping load growth. Used by infrastructure PE, sovereign desks, IPPs, and hyperscaler energy teams.

  • Bankable IC memos with sensitivity bands, not point estimates
  • NEM and ERCOT active; Gulf and PJM extending in 2026
  • Lattice (market) and Reservoir (asset dispatch) calibrated per region
  • Senior-analyst signature on every memo
PRODUCT.02 // QUANTITATIVE DECISION ENGINE · SUPPLY CHAIN

Nexus

The quantitative decision engine for supply chain. Sourcing, compliance, and contract intelligence across BESS, solar, wind, and balance-of-plant equipment. Used by procurement leads, EPC originators, lender diligence groups, and project finance teams.

  • RFQ workflows priced against market benchmarks, not vendor lists
  • Live compliance matrix across UFLPA, IEC 62933, and GCC ICV
  • Sourcing (intake) and Contracts (drafting) calibrated per programme
  • Senior procurement-lead sign-off on every brief
03 / PRODUCTS AT A GLANCE

Same engine. Two surfaces.

Athena and Nexus run on the Felixfusion platform. Both engagements run on the same clock and the same human-in-the-loop discipline. The buyer profile and the surface area differ; the runtime does not.

AXIS
Athena
QUANTITATIVE DECISION ENGINE · ENERGY
Nexus
QUANTITATIVE DECISION ENGINE · SUPPLY CHAIN
Domain01Energy decision intelligenceEnergy supply chain and procurement
Primary buyers02Infrastructure PE, sovereign desks, IPPs, hyperscaler energy teamsProcurement leads, EPC originators, lender diligence, project finance
Surfaces03Utility-scale BESS, hybrid renewables, data-center power dealsRFQ workflows, sourcing, compliance, contract structuring
Internal engines04Lattice (market) · Reservoir (asset dispatch)Sourcing (intake) · Contracts (compliance + drafting)
Headline output05Bankable IC memo with sensitivity bandsCompliance brief with provenanced clause library
Regions06NEM · ERCOT active. Gulf and PJM extending 2026.NEM · ERCOT active. Gulf and ICV regimes extending 2026.
Pilot shape07Four to six week engagement against a single decisionFour to six week engagement against a single capture
04 / WHO USES THE PLATFORM

Two engines. Eight buyer profiles.

The platform serves the people who write checks and the people who write contracts. Every buyer profile below has a specific decision shape that maps to either Athena or Nexus.

ATHENA · ENERGY
  • Infrastructure PE & sovereign desks

    Multi-asset valuation, portfolio NAV refresh, transaction-pipeline sizing.

  • Hyperscaler energy teams

    AI campus power demand modelling, PPA + BESS + on-site generation portfolio comparison.

  • IPPs & developers

    Bankable revenue forecasts, FID-ready sensitivity, siting and capacity scoring.

  • Lenders & structured finance

    Collateral performance, covenant monitoring, drawdown gates, stress scenarios.

NEXUS · SUPPLY CHAIN
  • Procurement leads

    RFQ workflows for BESS, solar, wind, and balance-of-plant equipment.

  • EPC originators

    Contractor scoring, scope and schedule risk, performance-guarantee structures.

  • Lender diligence groups

    Supply-chain audits, UFLPA / IEC 62933 / ICV compliance, supplier risk register.

  • Project finance teams

    Procurement milestone tracking, lender approval gates, multi-asset scheduling.

The compliance matrix built itself faster than our analysts could read the RFP. We compressed a fortnight of mechanical work into an afternoon and kept the senior judgement calls exactly where they belong.

Director, Infrastructure PE GroupPilot conversation, 2026 Q1

SCENARIOS - Three scenarios the engine is built for.

We are a seed-stage company. The stories below are illustrative: they show the shape of work the engine is designed to handle, drawn from the kind of engagements our pilots target. They are not case studies of shipped customer outcomes.

SCENARIO.01 // CAPITAL ALLOCATION

A fund evaluating a renewable interconnection.

An infrastructure investor is evaluating a direct acquisition at the late stage of an interconnection queue. Their internal process for a DD pack of this scope typically runs six weeks. The seller's exclusivity window is shorter than that.

Constraint:The team needs a defensible DD pack covering interconnection risk, system impact study findings, curtailment exposure, and PPA credit quality, on a schedule their usual advisors cannot meet. The artifact has to be readable by the fund's investment committee.

Read the full scenario
SCENARIO.02 // PUBLIC PROCUREMENT

A contractor responding to a federal tender.

A prime contractor is responding to a federal RFQ with a compressed submission window. Scope spans multiple technical volumes, a long list of mandatory compliance clauses, and a pricing exhibit that has to reconcile against internal cost systems.

Constraint:Their existing capture process assumes a longer runway. Under a compressed clock the two outcomes are a no-bid or a bid where compliance gaps surface at the evaluator's desk. Neither is acceptable.

Read the full scenario
SCENARIO.03 // PIPELINE TRIAGE

A utility triaging a filings backlog.

A regional utility has accumulated a multi-year backlog of interconnection cluster-study filings it has not triaged. The backlog is blocking buildable-site planning across its territories.

Constraint:Each filing runs 400 to 1,200 pages. A single internal planner can clear one filing every two to three days. There is no internal headcount to clear the backlog without either hiring or a tooling intervention.

Read the full scenario
PLATFORM

The runtime, the surfaces, and the artifacts.

Felixfusion is a single multi-agent runtime. Two product surfaces sit on top of the runtime. Each one produces an artifact a senior counterpart will sign their name to.

LAYER · INPUTSREGULATORY FILINGSGRID / MARKET DATARFP + TENDER PACKSASSET PERFORMANCELAYER · RUNTIMEFelixfusion enginev0.4 · MULTI-AGENT RUNTIMEPARSERead filingsSTATEHold contextSCOREQuantify riskDRAFTCompose memoAPPROVEHuman signsSURFACE · ATHENAQuantitative decision engine for energyLATTICE · RESERVOIRSURFACE · NEXUSQuantitative decision engine for supply chainSOURCING · CONTRACTSOUT · BANKABLE IC MEMOOUT · COMPLIANCE BRIEF
SCHEMATIC · ENGINE LAYERS NOT EXHAUSTIVE
03 / REGIONS

Where the engine runs, and where it is extending.

Athena's two models, Lattice and Reservoir, are calibrated against three market regimes. The depth and readiness of each region reflects how much of the local ruleset is carried in the runtime today.

ACTIVE

National Electricity Market (AU)

Primary depth. Regional reference pricing, NEM rules in force, loss factors against the published MLF and DLF schedules, AEMO dispatch emulation. First pilot shape for capital allocators.

ACTIVE

ERCOT (US)

Primary US surface for data-center-aligned BESS and hybrid renewables in 2026. Nodal LMPs in force, large-load queue intelligence carried in the runtime, four co-location structures priced.

EXTENDING 2026

Gulf Cooperation Council

GCC initial contact phase: UAE, Saudi Arabia, Oman capacity tendering and bilateral offtake structures are the first surfaces mapped. Pilot-ready in 2026.

GCC posture and surfaces
EXTENDING 2026

PJM (US)

Initial scope: capacity-market clearing, FTR-priced basis to data-center nodes, behind-the-meter regulatory reads on the Amazon to Talen precedent.

ROADMAP

Further ISOs, UK, EU

Additional North American ISOs (MISO, CAISO, SPP), the UK Balancing Mechanism, and selected EU national markets. Methodology extensions scoped; ruleset ownership is next.

BLOG - Notes from the build.

Long-form pieces from the Felixfusion team. Operator-grade analysis on the energy decisions that move the most capital.

RESEARCH12 MIN

The AI campus power stack in 2026.

Hyperscaler PPAs, BESS revenue stacks, and the queue mechanics shaping where the next 30 GW of compute lands.

Read the post
ANALYSIS8 MIN

The ERCOT LLIS mirage.

Two hundred and thirty-three gigawatts of intent. Roughly thirty are credible by 2028. The procurement consequences.

Read the post
BRIEF6 MIN

UFLPA and the BESS supply chain.

What the Uyghur Forced Labor Prevention Act enforcement update means for utility-scale BESS sourcing in 2026.

Read the post
04 / SECURITY

Designed to pass security review.

ISOLATION
Each pilot runs in its own sandboxed execution context. No shared state, no cross-tenant data paths.
ENCRYPTION
TLS 1.2+ in transit. AES-256 at rest. Key management delegated to the cloud provider's KMS, with customer-managed keys supported in scoped deployments.
CONTROLS
Design informed by NIST 800-53 and SOC 2 Type II frameworks. We are not SOC 2 certified today. Formal audit is on our 2027 roadmap.
POSTURE: PILOT-READY · FORMAL AUDIT: PLANNED 2027Request enterprise pilot