Consulting
Trust your materials databefore you build on it.
Materials Decoded is the consulting side of this publication. I help R&D teams and materials-AI startups audit the data they rely on, build the pipelines that feed their models, and ship AI tools for real materials workflows.
by Ibtisam Ahmed Khan· materials engineer working in AI + data
Who I work with
Materials-AI startups
Building models and tools on public and proprietary materials data, and needing it to be trustworthy.
R&D and lab teams
Sitting on scattered experimental and simulation data that could be feeding better decisions.
Investors and diligence
Needing an independent read on the data and models behind a materials or deep-tech deal.
How I can help
fixed fee, scoped up frontMaterials Data Audit
Where your data disagrees, and which numbers to trust.
The problem
Teams train models on values pulled from Materials Project, OQMD and internal records that quietly disagree with each other. Build on the wrong number and the error propagates silently.
What you get
A report on where the sources conflict, which values to trust, what is physically wrong, and exactly what to fix first.
Fixed fee, roughly one to two weeks.
Materials Data Pipeline
Scattered sources into one clean, modelled dataset.
The problem
Your data is spread across APIs, spreadsheets and papers, and every project starts by wrangling it again from scratch.
What you get
A reproducible pipeline that fetches, cleans and harmonises your sources into a dataset your team can actually model on, with the code and documentation handed over.
Scoped project, fixed fee agreed up front.
AI for Materials
A working agent or tool for a real workflow.
The problem
You want AI inside a materials workflow, not another toy demo that falls over on real data.
What you get
A runnable tool grounded in real data: literature screening, candidate screening, or characterisation triage, with guardrails around anything expensive or irreversible.
Scoped project, fixed fee.
Need something ongoing? I also take a small number of advisory retainers, a few days a month to keep a materials-AI effort on track. Ask on the call.
Why Materials Decoded
everything here, I have built and runI am a materials engineer who moved into data and AI, and I keep noticing the same thing: the data underneath materials science is far less trustworthy than the models built on top of it assume. Every service I offer, I have already built and run in the open.
A live Materials Project auditor
Flags where a public database contradicts itself, deployed and running.
Multi-source data pipelines
Fetch, harmonise and validate across Materials Project, OQMD and more.
A series of materials AI agents
Literature, failure diagnosis and discovery agents, each open on GitHub and Hugging Face.
A cross-database trust dataset
Measuring exactly how much two standard databases disagree, and why it matters.
See it: the writing ·GitHub ·Hugging Face
The house rule is honesty. If your data is fine, I will tell you and keep the invoice short. If a fix lowers a model's score, you get the lower number.
How it works
- 01
Scope call
A short call to understand the problem and whether I am the right fit. No charge.
- 02
Fixed proposal
A written scope, deliverable and fixed fee, so there are no surprises.
- 03
The work
I build, with checkpoints along the way so you see progress, not a black box.
- 04
Handover
A clear deliverable you own outright: the report, the pipeline, or the tool, with its code.
Have a materials data problem?
Tell me what you are building and what you are unsure about. If I can help, we will scope it. If I cannot, I will say so.
Book a call ->