Skip to content

Celaron — an independent AI company

AI for the work that has to be right.

We build software for people who measure things: materials researchers who cannot afford a fabricated number, and oncology teams who cannot afford a missed variant. Three platforms are live today.

Start free on MatflowSee the platforms

OPEN SIGNUP · SELF-HOSTABLE · EVERY NUMBER CARRIES AN EVIDENCE CLASS

matflow.celaron.com — Optimization Studio
Matflow Optimization Studio showing a Pareto frontier of predicted catalyst candidates with conformal intervals
Real Matflow output: a Pareto frontier of predicted candidates, each with conformal intervals and an evidence class. Illustrative demo data; labels are live.

0.41–0.95

Held-out R² across Matflow's public benchmark datasets

matflow.celaron.com/benchmarks

510,504

Cancer variants in CAPRI's live catalogue

capri.celaron.com/api/stats · Oct 2026

36

Workflow modules, from ingestion to techno-economics

Matflow feature index

3.4×

Lower forgetting in a multi-seed CMM audit

FM 0.054 → 0.016 · 27 runs

Numbers above are read from the live products and their public benchmark pages. Where a number cannot be verified, we do not publish it.

The platforms

Three products, one discipline.

Each platform is engineered the same way: transparent evidence, real engines, and data you can export or self-host.
All platforms

01 — Scientific R&D

Matflow

Physics-validated AI for batteries, formulations, and materials.

Experimental rows in, ranked candidates out. Matflow turns sparse lab data into working models, cost-aware candidate rankings, and publication-grade dossiers — with every number traceable to its source run, engine, and evidence class.

  • Upload CSV, Excel, PDF, or instrument files — XRD, GC/MS, FTIR, NMR, battery cyclers — with automatic column mapping, unit extraction, and QC review.
  • Generative augmentation (Neural GAN, Hybrid GAN + copula, or Statistical) audited by 20+ quality, privacy, and distribution metrics before it reaches a model.
  • Five-model ensembles with SHAP/LIME explanations, split-conformal prediction intervals, and explicit extrapolation flags.
  • Multi-objective optimization (adaptive and evolutionary search) with TOPSIS ranking and true Pareto fronts under 100% sum constraints.
Start free on MatflowMatflow details

0.41–0.95

Held-out R² across public benchmark datasets

0.95

Leave-family R² on the bundled demo corpus

6

Public gold suites, reproducible end to end

36

Workflow modules across data, modelling, and lab automation

02 — Precision Oncology

CAPRI

Evidence-ranked variant interpretation, in one workbench.

Cancer Actionability & Precision Response Interpreter. CAPRI unifies twenty-nine biomedical sources into a single workbench — normalized variants, AMP/ASCO/CAP-style Tier I–IV actionability, structure-aware drug context, cohort analytics, and de-identified tumor-board summaries.

  • Unified search across genes, variants, drugs, and diseases, with transparent evidence ranking.
  • Upload VCF, MAF, or TSV cohorts; variants are validated, normalized, and quality-checked before analysis.
  • Curated clinical evidence from CIViC and CGI with AMP-style Tier I–IV actionability and A–E evidence levels.
  • Structure context: PDB mappings, residue-level annotation, pocket structures, and docking jobs with pose SDF output.
Create a CAPRI accountCAPRI details

510,504

Variants in the live catalogue

1,084

Genes in the live catalogue

29

Integrated biomedical sources

2,436

Drugs with target and sensitivity context

03 — Private AI chat

Celaron Chat

A private AI workspace you actually control.

Celaron Chat is a self-hosted workspace for everyday AI work — conversations, documents, and images on infrastructure you control, with the model options you choose.

Open Celaron Chat
  • Self-hosted and private by default — conversations stay under your control.
  • Email or Google sign-in with per-user access control.
  • Workspaces for documents, images, and everyday AI tasks.
  • Built on the open-source Open WebUI stack (v0.8.12) — inspect it, extend it, or run it yourself.

The evidence habit

Nothing fake is ever presented as real.

Every number our platforms produce carries an evidence class and a link back to its source run or engine. It is a small design decision with a large consequence: you can tell at a glance what was measured, what was computed, and what is still a guess.

We hold our own marketing to the same rule. If we cannot verify a claim, it does not go on this website.

MEASURED

Raw experimental or instrument result.

COMPUTED

Output of a real engine run — PyBaMM, M3GNet, pycalphad.

PREDICTED

Registered model output with calibrated uncertainty.

EXTRACTED

Pulled from literature, flagged for review.

HYPOTHESIS

A generative proposal, never presented as fact.

DEMO

Synthetic or illustrative — never billed, never disguised.

How we build

Four rules we do not break.

The AI industry is full of claims. We would rather be boring and correct than exciting and wrong.

01

Evidence over hype

Every claim our platforms make carries a provenance trail. Numbers are labelled MEASURED, COMPUTED, PREDICTED, EXTRACTED, HYPOTHESIS, or DEMO — and nothing fake is ever presented as real. That rule applies to our own marketing, too.

02

Open by default

We build on open standards and open-source science engines. Data is exportable, models are portable, and the platforms can be self-hosted — because the people doing the work should own the work.

03

Human in the loop

AI accelerates decisions; humans make them. Our platforms are designed with approval gates, audit trails, and safety interlocks at the steps where being wrong is expensive.

04

Small data is the real world

Most scientific teams have tens of experiments, not millions of rows. We build for sparse, expensive, hard-won data instead of pretending everyone has internet-scale datasets.

Research

Published, not promised.

Celaron grew out of research internships at QCRI and QEERI. The work is public — read it and judge for yourself.
All research

2026

Acquisition, Retention, and Fisher-Proxy Scale in Continual Model Merging

Working manuscript · independent audit of Phan et al. (2025), arXiv:2509.23592

Audit code & results

2025

CatForge — open catalyst design platform

QCRI · first author on the system paper (manuscript in submission)

Visit CatForge

2026

C2CAD-BENCH — benchmarking LLM spatial reasoning

QEERI internship project · contributor to implementation, experiments, evaluation, and manuscript (not founder)

View the benchmark
Bassil Elshenawy, Founder & CEO of Celaron

From the founder

“I kept meeting brilliant people whose tools lied to them. Celaron is my attempt to build the opposite: software that shows its work.”

Bassil Elshenawy — Founder & CEO. Previously a GenAI research intern at QCRI and QEERI; founding contributor of the CatForge catalyst-design platform.

View the portfolioLinkedIn

Start with Matflow — or tell us what you are working on.

Free accounts include every Matflow module, academic use is free, and partnerships start with a conversation. Either way, a human answers.

Start free on Matflowbassil@celaron.com