01
Massively Parallel Synthesis
scale 250M
Material instances per chip
100K
Unique compositions per chip
scope 52+
Elements, arbitrarily mixed
quality 2,500
Replicates per material
The materials
we need
don’t exist yet.
Mattiq generates the ground truth that turns AI prediction into reality.
02 / 06 Physical-world materials data bottleneck
Materials AI
has a reality
problem.
We ground it
in experiment.
Why today’s materials AI falls short
-
Today’s models are trained largely on simulation — and inherit every approximation and bias baked into it.
-
Experimental datasets cover only a tiny fraction of possible compositions, structures, and conditions — and almost no performance — forcing models to extrapolate beyond measured reality.
-
A predicted material can promise record performance. On paper. Simulations stop short of process and proof. Mattiq delivers both.
-
Prediction without experimental data is lacking. Current AI models are limited by a lack of quality training data. Mattiq is building the world’s fastest and most comprehensive ground-truth data engine for materials discovery.
03 / 06 Megalibrary Data Engine
Ground truth at AI scale.
Mattiq brings genomics-scale experimentation to inorganic materials through its chip-based Megalibrary platform.
By scaling ground-truth data generation, Mattiq is laying the foundation for the world’s largest experimental materials dataset—and a radically faster learning loop between prediction and experiment.
02
Multimodal Characterization
scope 20+
Measurable properties / material
size · composition · crystal structure · mechanical · magnetic · electrical · optical · chemical · thermal quality 25+
Replicas measured / material
throughput 10K+
Materials measured / day / instrument
03
Unified Data Record
IDENTITY
Every material retains its composition and encoded position.
PROCESS
Inputs, conditions and complete synthesis history are preserved.
OUTCOME
Structure and functional properties remain co-registered.
MACHINE-READABLE
Everything resolves into one unified training record.
04
AI Grounded in Physical Reality
MEASURED TRUTH
Models learn from physical experiments—not simulation alone.
COMPLETE SIGNAL
Successes, failures and replicates all become training data.
CLOSED LOOP
Every prediction can be tested and returned to the model.
COMPOUNDING LEARNING
More experimental evidence reduces error and broadens generalization.
04 / 07 Modular by Design
Ground truth at the scale AI needs to generalize.
Dimensions — composition, processing, structure, properties, operation.
One dataset. Many material frontiers.
01
The challenge
- Constraint
- PEM electrolysis depends on scarce iridium (Ir) electrocatalyst operating under harsh acidic conditions.
- Objective
- Use less without sacrificing activity or durability.
- Barrier
- Millions of possible compositions—beyond the reach of serial experimentation.
- ~7 t
- Annual global
Ir production - 30 yrs
- Years of incremental OER progress
02
The search
IrM1M2
IrM3M4
IrM5M6
IrM7M8
- 14
- Trimetallic design
spaces synthesized - 1M
- Candidates
screened - 6 wks
- From campaign
launch to leads
03
The proof
Leads were scaled from chip to gram quantities and validated in PEM electrolyzers for 1,000+ hours, in-house and with industrial partners.
Mattiq and Heraeus Partner to Advance Electrocatalyst Innovation
- 500+
- Promising candidates
discovered - 3
- Scaled and validated
in PEM electrolyzers - 60%
- Reduction in Ir use
and catalyst cost
57,000 Estimated years to screen the same candidates using traditional methods one at a time.
Electrochemistry
Map activity and durability across million-candidate catalyst spaces — for hydrogen, fuel cells, and CO₂ conversion.
Magnets
Screen compositions for coercivity, anisotropy, and thermal stability — from rare-earth-free magnets for traction motors, robotics, and defense to recording media for data storage.
Batteries
Link composition and processing to ionic transport, stability, and interfacial behavior — ground truth for electrodes and electrolytes.
Memory
Screen switching, retention, endurance, and variability across thousands of replicas per composition — materials for resistive, phase-change, and ferroelectric memory.
Superconducting Materials
Map phase formation, structure, and doping across vast multinary spaces — candidate superconductors for fusion magnets, power, and high-field systems.
Semiconductor Materials
Screen leakage, resistivity, and stability across dielectrics, oxide semiconductors, and interconnect metals — materials for gate stacks and advanced nodes.
From industrial R&D to AI and autonomous laboratories, better experimental ground truth makes every layer more predictive.
05 / 06 The Mattiq Team
Materials discovery, delivered.
Born from Professor Chad Mirkin’s pioneering Megalibrary technology, Mattiq is building the missing experimental layer for materials intelligence. Every era has been limited by its materials. Our ambition is to make ours the first that isn’t.
Founders, board & leadership
Prof. Chad Mirkin
Scientific Founder & Director
Ben Schlatka
CEO & Director
Carmichael Roberts
Director & Investor, Material Impact
Quinten Stevens
Director & Investor, CS Venture
Dr. Andrey Ivankin
Co-Founder & CTO
Duane Dickson
Chief Commercial Officer
Alex Mantis
Director of Data & AI
Dr. Carolin Wahl
Director of R&D
Backed and supported by
06 / 06 Start a conversation
Join the mission Shape the future of
Shape the future of
the material world.
Partner. Advise. Invest. Build with us.