Process evidence · pharma manufacturing

COAs tell you what probably happened. I show you what actually did.

Custom-designed sensor kits deployed at your supplier's production site, paired with software that turns raw process signal into reports your CMC team can actually use. Independent. Physically grounded. Verifiable.

R-100 · EDGE NODE Live
T-102 reactor temp
68.40°C
pH-104 in-line probe
7.08
I-101 stirrer current
2.14A
P-103 vessel pressure
1.020bar
SHA-256 · CHAIN-OF-CUSTODY block #00000

Your contract manufacturer passes every GMP audit. Your API still shows batch-to-batch variability that nobody can fully explain.

Today, what you know about your supplier's process comes through a paper trail — batch records, certificates of analysis, quarterly reports. By the time a deviation surfaces in your CMC package, the reaction is already six weeks in the past.

Symptom

"Root cause undetermined."

Every CMC team has read this line in a deviation report. Without process-level data captured at the time of the event, root cause analysis becomes archaeology.

Symptom

Certificates outrun reality.

The COA tells you the batch met spec on the sampled aliquot. It doesn't tell you what happened during hours 3 through 7 of the reaction — or whether the next batch will behave the same way.

Symptom

The trust-based ceiling.

"The CMO said so" is the highest level of confidence you can currently purchase. For clinical programs where a single failed batch delays an IND filing, that ceiling is too low.

Symptom

Clinical feedback loops are too slow.

By the time a batch-to-batch variability issue reaches your clinical sample, the affected material has been consumed, the process drift has already completed, and you're investigating a ghost.

Instrument the process. Make the evidence physical.

Every engagement is custom-designed around the specific chemistry, reactor configuration, and quality-relevant parameters of your product. The deployment typically spans four steps.

01
Parameter map

Identify the process signals that actually govern your product quality.

I work with your CMC team to build a quality-by-design parameter map specific to your chemistry. For an exothermic reaction with moisture-sensitive intermediates, this might be exotherm profile and moisture; for a high-pressure catalytic step, pressure-drop rate and catalyst load. The outcome is a short list of physical signals that — if measured continuously — would have caught every deviation in your deviation history.

02
Deploy

A custom-designed sensor kit, installed at your supplier's site.

I ship a sensor kit engineered against your parameter map — temperature, pH, mechanical load, pressure, and whatever else your chemistry demands. Each kit uses industrial-grade, non-invasive sensing matched to the specific reactor and process. Physical install is coordinated with your supplier; electrical and data isolation ensure I never touch their control system.

03
Capture

Continuous capture with tamper-evident integrity from the moment the data is born.

Sensor data is sampled at process-relevant rates and written to a hardware-signed log at the point of measurement. Every record is timestamped and cryptographically chained — so that the evidence is provably the same evidence that was produced at time of reaction, not a reconstruction assembled after the fact.

04
Deliver

Reports your CMC team can drop into the package — not raw data you have to decode.

After every batch, my software translates raw sensor streams into structured reports: batch-to-batch comparison, deviation flagging against your spec, integrity attestation, and narrative summaries written in the language of CMC documentation. If something drifted, you know before the COA arrives — not six weeks after your clinical sample fails spec.

Consumer-hardware reliability discipline, applied to pharma supply chains.

AtomTruth exists because the gap between what pharma manufacturing can measure and what consumer hardware already measures is now almost a decade wide.

At Apple, a 0.1% defect rate meant millions of failed devices. The only way to survive that math was to instrument every supplier process down to the shift and the machine — to treat quality as a measurable, traceable, enforceable signal rather than a trust-based claim. That discipline is what I'm bringing to the APIs and intermediates going into clinical trials.

I'm building this for North American biotechs. If supply-side variability has ever cost your program time, data, or a clinical sample — that's the problem I'd like to hear about.

Background
Founder
Qiran Xiao Mountain View, CA
Prior
Hardware Reliability Apple · Google (Pixel)
Training
Ph.D., Materials Science Stanford University
Shipped
20+ hardware programs shipped Zero field escalations across portfolio

If your last deviation investigation ended with "root cause undetermined" — let's talk.

qxiao@atomtruth.com

20-minute conversations. No pitch deck. I come with three specific questions about your current process-evidence workflow, and share what I'm learning across other conversations.