From scientific discovery to industrial precision.

Aikemics packages 13+ years of computational simulations into hybrid physics and AI models that cover every battery manufacturing step from the slurry to the cell. Gigafactories and labs use them to decide faster, on infrastructure they control.

Book a call →See the platform
13+
Years of research maturity
100+
Validated physics & AI models
2100+
Manufacturing scenarios reproduced
16M€
Cumulative R&D funding behind the platform
Digital twin of a battery electrode coating line, rendered in blue with simulation dashboards overlaid
The problem we solve

Battery manufacturing is still trial-and-error.

A new recipe takes months of physical experimentation to qualify on the line. A new format or line transfer can take years. Material losses during ramp-up run up to 30%. At steady state, a 30–40 GWh plant can burn around 150 M€ a year in scrap. Qualifying a new chemistry can take a decade.

The alternative is predictive engineering. We model the process with mechanistic physics. Where the simulation is too slow for production decisions, AI surrogates take over.

01
Trial-and-error costs millions on every gigafactory ramp-up. Material losses can reach 30% before yield stabilises.
02
Process windows are brittle. A format change, a line transfer or a chemistry swap can break them, and the recipe needs weeks of physical re-tuning before it holds again.
03
Data-driven ML alone is unreliable on the sparse, drifting data a line actually produces. Mechanistic physics is the honest answer, but it's too slow for real-time decisions. We pair the two so the physics drives the predictions and the AI surrogates make them fast enough for the line.
04
Engineering knowledge lives in spreadsheets and lab notebooks. Nothing systematic captures or versions it, so when someone leaves the team the knowledge goes with them.
Why Aikemics — five differences

Built for industrial-grade performance and everyday usability.

Our platform is opinionated. These five principles are what make hybrid intelligence actually deployable on a line rather than impressive in a paper.

01 / 05

High fidelity

Physics-driven accuracy across scales. Validated against 2100+ manufacturing scenarios.

02 / 05

Trustable

Every prediction ships with provenance: the model that produced it, its audit trail, its uncertainty band. The platform is built on 185+ peer-reviewed publications, not opaque regressions.

03 / 05

Real-time

Insights when decisions happen. Surrogate ML models run in milliseconds on the line. Heavy physics jobs run in the background on HPC or cloud.

04 / 05

Flexible

Chemistries (NMC, LFP, Si-anodes, sodium-ion, solid-state), formats (cylindrical, pouch, prismatic) and pilot designs all run on one stack.

05 / 05

User-friendly

Clean UX with actionable guidance. Validated playbooks surface where engineers actually work. No PhD in electrochemistry or material processing required.

What it brings — by use case

Fewer reworks. Faster time-to-production.

One stack, three audiences. The platform fits a process engineer tuning a slot-die head as cleanly as a materials scientist screening binders or a lab lead versioning experiments.

Aerial view of a battery gigafactory with a solar-panelled roof, ringed by farmland
INDUSTRY · MANUFACTURING

Battery manufacturers & gigafactories

  • Calibrate process windows to improve reproducibility
  • Shorten ramp-up experimentation by up to 10× before serial production
  • Recover ~30 M€/year for every percentage point of scrap avoided
  • Scale recipes across lines, formats and chemistries
Explore use cases →
Close-up of a polished stainless-steel industrial slurry mixing vessel
INDUSTRY · SUPPLIERS

Equipment & materials suppliers

  • Co-optimize equipment settings and material specs
  • Provide model-informed application notes to customers
  • De-risk new formulations and process ranges before customer trials
Explore use cases →
3D simulation of a battery electrode microstructure on labelled X, Y and Z axes
INDUSTRY · RESEARCH

RTOs & advanced labs

  • Unify experimental and simulation workflows in one versioned environment
  • Move from one-off studies to repeatable, sharable, replayable models
  • Train teams on predictive engineering at scale
Explore use cases →
Based on research

Thirteen years of physics.
One deployable platform.

Aikemics is the spinout of Prof. Alejandro A. Franco's research group at LRCS (Laboratoire de Réactivité et Chimie des Solides, CNRS UMR 7314) at Université de Picardie Jules Verne. Prof. Franco's group is one of the world's most-cited research groups in computational modeling and AI. Every module in the platform traces back to peer-reviewed models that were validated and stress-tested before they ever touched a customer line. The platform is the bridge from scientific discovery to the industrial precision a production line demands. An exclusive licensing agreement with CNRS/LRCS is in final negotiation.

Recent representative work from our CSO
2025ECS M. Stanley Whittingham Mid-Career Award — for extraordinary contributions to intelligent battery manufacturingElectrochemical Society
2024Digital twins for electrode manufacturing: a multi-physics + ML frameworkAdvanced Energy Materials
2023ARTISTIC — ERC-funded digital model of the Li-ion battery manufacturing processEuropean Research Council
13+
years of computational modeling and AI research behind every model
185+
peer-reviewed publications informing our scientific backbone
2
ERC grants (Consolidator + Proof-of-Concept) won by our CSO
16 M€
cumulative R&D funding behind the platform's foundational science
Common questions

Questions we hear most often.

Anything missing? Send us your question →

No. The platform supports operators, process engineers and scientists by codifying validated knowledge and surfacing it where decisions get made. Your team still owns the call. The platform just makes the trade-offs visible and auditable instead of tacit.
Yes, and we believe this is essential. Aikemics is built around a hybrid architecture: validated model libraries in a secure cloud and sensitive computation deployed on your servers or HPC. The platform is available as managed cloud, private VPC or fully air-gapped on-premise, depending on your data sovereignty and IP requirements.
Recipes, process data and IP never leave your environment in our on-premise and VPC modes. Models are versioned and signed. We are GDPR / RGPD compliant by design, and the hybrid architecture was built specifically for European sovereignty constraints.
Today: NMC, LFP, Si-doped graphite anodes; cylindrical (18650, 21700), pouch and prismatic formats. Sodium-ion and solid-state are in active development with research partnerships
Aikemics Studio ships progressive product families: CO-SIMU (finite-element simulators, available today); ML-PREDICT & ML-REVERSE (predictive and inverse ML models, Q4 2026); ML-INSIGHT (recovery of hidden process variables, Q2 2027); and ML-FEEDBACK (supervised autonomous adjustment, Q1 2028).
A scoped problem, access to one process step's historical data, and a few weeks of joint definition. We come back within a week of first contact with a feasibility note and a path to a first deployable model.
Let's build

Bring us a scoped battery-manufacturing problem.

We come back within a week with a feasibility note, a proposed model and a path to first deployment.

Book a call →Send us a problem