The operating system for the battery & energy lifecycle.
Site development, simulation, diagnostics and compliance — from the first parcel of land to second life, on one project record.
In production with industry partners.
Axioma Intelligence is deployed under NDA with manufacturers, operators and insurers across the battery value chain. Our first two European deployment partners:
Two partners, two continents, live across development, operations and compliance.
Customer name withheld during commercial onboarding. Named references available on request.
The industry produces enormous amounts of data and moves almost none of it between the people who need it. We are a small team of engineers, scientists and operators building one platform to move it. Data · Action · Assurance.
Evidence produced once, used everywhere it is needed.
Most battery data is produced once and used once. A cell test settles a question in a lab report. A cost basis wins a board paper — and then each one stops. The next person who needs that number derives their own version, from worse inputs, and defends it from scratch.
We build for the opposite. One record, many readers — developer, EPC, owner, underwriter, lender, recycler: the work that establishes a fact happens once, and every discipline downstream reads it in its own language.
Every figure arrives with its source and its grade attached: screening estimates labelled as such, models that name themselves. A trail built for a regulator or a lender's adviser.
Anyone can build a better site designer, a better underwriting model or a better passport form. The hard part is making the output of each stage usable by the next one — without retyping it, and without losing where it came from. That is the product. The modules are how it is delivered.
Example 1 of 5: One costed site design, read by An EPC scope package, A tender RFQ, The finance capex basis.
Three teams, three deadlines, three different questions — and one piece of evidence between them. Building the platform so that happens mechanically, rather than by someone remembering to send an email, is the whole design.
These five are seams that carry data today. Most of the chain's seams do not yet — a passport still does not price its own second life, and an operating history does not yet grade its own inventory. We would rather name the ones that work than imply the whole chain is joined up.
Screening tools that show their working
Everything here narrows the field and tells you how good each number is. Detailed engineering studies, ground surveys and professional underwriting judgement still make the final call — and we would rather tell you the grade of a number than let you assume it is better than it is. Wherever a figure on this site is screening-grade, it says so.
One platform across the battery value chain.
Eight capabilities, wired to one source of truth — from the data layer to the compliance filing. Every one is a cell of the same pack.
Unified data layer
BMS, lab, manufacturing, and field assets in one schema.
Physics & AI simulation
BattMo, ECM and ageing models, calibrated to a customer's own data.
Manufacturing insights
Defect analytics and yield, linked to field performance.
Real-time diagnostics
SoH, thermal risk, and degradation for every cell and site.
Warranty & lifetime intelligence
RUL forecasting, warranty shield, and degradation-mode diagnostics.
Risk analytics
Underwriting, claims prediction, and fleet benchmarks.
Second-life & circular
Repurposing, SoH-graded inventory, recycling logistics.
EU Battery Passport
Digital product passports, structured to the Feb 2027 mandate.
NVIDIA selected Axioma Intelligence for its Innovation Lab: a 60-day programme with dedicated time on an eight-GPU NVIDIA H100 machine — the class of computer used to train today's leading AI systems. The programme's original use case, assessing the risk of shipping lithium battery cargo, was developed with an insurance design partner. The first fully governed runs completed in July 2026. The current sprint runs through August 2026, across three areas: battery materials, shipping safety and site intelligence.
Eight models run or are in build on the lab machine. In plain words, what each one does:
Decides whether a battery shipment should be accepted, referred to an underwriter, or excluded — and shows which facts drove the decision. Trained on a million example shipments constructed from public incident records and the official rules for transporting batteries.
So far: Served in the platform today, sub-40 ms in validation, with the model's actual reasons shown per decision. Training labels are still rule-derived until partner claims data lands.
Simulates many years of possible battery-fleet losses to show how bad a bad year gets, and how often — the kind of model insurers use to understand hurricane exposure, purpose-built for portfolios of battery storage sites.
So far: A 50-site, 100,000-year methodology run is archived. The full portfolio run on the lab machine is scheduled.
Predicts a battery cell's full discharge curve in milliseconds, where the physics simulation it replaces takes minutes. Built for our interactive materials design tool, and not yet connected to it.
So far: Trained on a 54,896-solve physics sweep and promoted for screening use on five chemistries after per-chemistry review (August 2026). It is not serving today: it needs a GPU host to run on.
Predicts cycle life, energy density and fade rate straight from a cell recipe, so a materials engineer can screen thousands of designs before running one simulation.
So far: Built on the same archived simulation campaign; each run carries its inputs and checksums.
Fits fade curves to real laboratory cells, so the platform's ageing assumptions are anchored to measured cells rather than a rule of thumb.
So far: Fitted to 32 NASA laboratory cells; fits below the platform's quality floor are excluded rather than smoothed.
Works out when a grid battery should charge and discharge to earn the most it could actually have earned across several electricity markets at once, obeying every physical and contractual limit. Its screening estimate of what a site could earn feeds the Project development flow below.
So far: A two-market scenario matrix runs in the platform today; the portfolio-scale run is in progress on the lab machine.
Scores and ranks candidate sites for a grid battery against land, demand and grid-connection evidence, so developers see early whether a site is worth pursuing — before commissioning full studies.
So far: Screening candidate sites inside the Project Development flow today.
Maps which pieces of a project's evidence actually connect to each other — and shows the gaps where they do not.
So far: The obligations record it maps landed in August 2026; the graph itself is the next build.
The July–August 2026 sprint scales up the battery simulation campaign and deepens the terrain and market models behind site intelligence. It also brings in NVIDIA BioNeMo molecular models, to help screen the chemical ingredients of new electrolytes.
Every pack sold into the EU will need a passport with a verifiable supply-chain thread behind it. That thread has to be built while the pack is being built, not filled in at the end.
From a parcel file to a costed site concept, in one sitting.
One flow, one project record. Battery storage projects usually live across a dozen disconnected tools; here every step reads from and writes to the same project, so nothing gets lost between teams and far more candidates can be screened before survey and engineering budget is committed. The flow can also draw on evidence established earlier in the supply chain — a cell test, a passport, a shipment history — rather than starting cold.
Import or draw the site
A parcel file, or a boundary drawn straight onto the map. Designs round-trip losslessly — what goes in is what comes out.
Screen the land
Terrain analysis from satellite elevation data — slope, buildable area, how much earth would need moving — plus flood, planning and grid constraints from public datasets. Each one licence-captured.
Lay out the plant
Batteries, inverters and transformers placed from a vendor component library at true physical dimensions. The design previews in 3D.
Cost and revenue screening
Indicative build costs read off the layout as drawn, and a screening estimate of what the site could earn across electricity markets.
Hand off to finance & EPC
The costed concept flows into the finance business case and EPC tendering — same project record, no re-typing.
A model earns its place two ways: it removes a wait, or it changes a decision.
A model that does neither is decoration. Those are different mechanisms, and we argue them separately: each of the five worked examples says which of the two it does, and what is in the platform today at what grade. We do not publish percentages or hours-saved figures we have not measured, and our design partners stay anonymous while commercial onboarding is under way.
A wait removed
A wait for compute, a wait for a specialist, or a wait for a document to be read. The claim is about elapsed time to a decision of the same quality — not about cutting corners to get there sooner.
The clearest example of a wait removed is the site-screening flow above — a parcel file to a costed concept in one sitting.A decision changed
Either it searches a space too large to search by hand, or it sees a signal nobody reading a spec sheet would catch, or it makes a judgement consistent where it used to depend on who ran it.
A capability appears in the platform only when it is real, labelled with its grade, and falls back honestly. A model never quietly replaces the deterministic calculation it sits beside — that stays the number of record, and the model's answer is kept next to it. When a model is unavailable the deterministic result stands in its place, not a different class of answer wearing the same badge.
27 elements, one platform.
Every module is a building block, and they share one data graph. Hover any element for what it does and what it produces.
BESS investment workspace — pipeline, technical due diligence, revenue-stack modelling, and post-COD performance tracking.
- Pipeline tracker
- Revenue-stack models
- Performance vs. base case
Terrain analysis from satellite elevation data — slope, buildable area, earthworks — plus flood, planning and grid constraints from around 30 public datasets. Candidate sites are screened before the survey budget is spent.
- Terrain & slope screening
- Constraint overlays
- Ranked candidate sites
A site boundary imported or drawn, a plant laid out from vendor components at true physical dimensions, a 3D preview, and a costed concept handed to finance and EPC. One project record, no re-typing.
- Plant layouts in 3D
- Costed site concepts
- Finance & EPC handoffs
BESS trading revenue modelled against GB market data: wholesale and balancing-mechanism screening grounded in Elexon settlement data, with scenario matrices across power, duration and degradation. Screening-grade by design.
- Elexon-grounded price screens
- Scenario matrices
- Project revenue case
Battery finance workflows — business cases built from costed site concepts, line-level CapEx traceability, portfolio dashboards, and reporting views for lenders and CFO teams.
- Business cases
- Cash-flow models
- Debt structure & min-DSCR tracking
Structured due-diligence reviews of battery assets and portfolios — technical, commercial and regulatory. Every step leaves an audit trail.
- DD memos
- Red-flag summaries
- Comparable benchmarks
Battery-specific risk analytics for insurers — Thermal Event Risk Score, claims prediction, fleet benchmarks, and underwriting decision support.
- Thermal risk score
- Claims probabilities
- Underwriting verdicts
A tender response from RFP to submission: RFPs ingested, requirements extracted, compliant responses drafted with AI, bids benchmarked against historical wins.
- Requirement extracts
- AI-drafted responses
- Bid score benchmarks
The EPC workspace reads the canonical project record, so a project's design and cost basis carry into procurement without re-typing. Contractors and owners negotiate from the same design-backed numbers the developer produced.
- EPC scope packages
- Design-backed cost basis
- Procurement handoff
Day-to-day BESS operations: SoC management, augmentation planning, contract compliance, availability KPIs, and revenue stack reporting.
- Availability KPIs
- Augmentation plans
- Revenue stack
Realistic BMS telemetry streams, generated under custom duty cycles. Pipelines, alarms and diagnostics get tested before deployment.
- Synthetic V/I/T traces
- Fault injection
- Replayable cycles
Solar generation forecasts, BESS dispatch optimisation, and energy-market trading signals — co-optimised against weather and price curves.
- 24-72h forecasts
- Dispatch schedules
- P&L attribution
One record for every battery asset across modules — solar farms, BESS sites, EV fleets, recycled packs — federated under a single schema.
- Unified asset list
- Cross-module IDs
- Status & geography
SoH and RUL assessment for every battery in an EV fleet — drives predictive maintenance, second-life eligibility, and remarketing pricing.
- Per-vehicle SoH
- Remarketing values
- 2nd-life eligibility
Live model replicas of fielded batteries — combine BMS telemetry with calibrated physics to predict thermal events and degradation in real time.
- Live SoH/SoC
- Anomaly alerts
- Predictive RUL
ADR / IATA / IMDG-compliant shipping workflows for batteries — UN38.3 documentation, hazard classification, route risk and chain of custody.
- Compliance packs
- Route risk scores
- Chain of custody
Battery R&D programmes, run end to end: experiments planned, samples tracked, cell test data captured, results linked back to the projects and hypotheses behind them.
- Experiment registry
- Sample lineage
- Project KPIs
Electrode and electrolyte materials screened by AI-assisted property prediction, before anything is committed to wet-lab synthesis.
- Property predictions
- Candidate ranking
- Synthesis briefs
Pseudo-2D electrochemical simulations on BattMo: SEI growth, lithium plating and capacity fade, for any cell design.
- P2D simulation outputs
- Degradation curves
- Design sensitivities
8 chemistry types and 6 test protocols — CC, CV, CC-CV, pulse, drive cycle, calendar ageing — with time-series voltage, SoC and temperature output. Screening-grade, for early cell comparison.
- Protocol simulations
- Chemistry comparisons
- CSV exports
Semi-empirical and physics-informed ageing models fitted to calendar and cycling data, then used to predict end-of-life under any operating profile.
- Ageing model fits
- EOL predictions
- Application simulations
Equivalent-circuit models — 1-RC and 2-RC Thevenin — built, fitted and simulated for fast online state estimation and SoC/SoH algorithms.
- OCV-SoC curves
- R/C parameter maps
- Voltage simulations
The federated data platform underneath everything — ingestion, quality monitoring, lineage, and governed access across all collections.
- Ingestion jobs
- Quality scores
- Lineage maps
Materials, carbon, lifecycle and supply-chain attestations, rolled up into one public-facing record. Structured to the EU's February 2027 mandate.
- Passport records
- Compliance score
- QR-linked data
End-of-first-life packs matched to second-life applications by SoH, chemistry and form factor, with circular value and recycling routes tracked alongside.
- SoH-graded inventory
- Application fit
- Circular passport
Warranty coverage, reserves and claim cycles in one place, with each claim linked back to its manufacturing lot, its field telemetry and the ageing predictions.
- Claim pipeline
- Reserve forecasts
- Lot trace-back
Cradle-to-grave carbon accounting for batteries: embodied emissions from materials, operational CO₂ avoided and created, and recycling credits.
- kgCO₂e per kWh
- Avoided emissions
- ESG-ready reports
On a small screen, tap a suite chip to see what each group covers.
Let's talk.
Your battery and energy data, on your assets, in 30 minutes.
Software, plus the team to make it land. Tell us what you are trying to decide.
info@axioma-intelligence.com