Founded 2022 · Denver, CO

Built by engineers who spent years inside fabs.

In 2021, Diana Muller was the process integration lead on a 7nm logic qualification line. Third shift was consistently producing escaped defect rates 3–4× higher than first shift on the same lot — identical equipment, identical recipe, different operator attention. The existing inspection tooling flagged everything and explained nothing. Operators re-inspected every amber die regardless of cause. Diana and Marcus Reyes left their respective roles in early 2022 to build a classification layer that actually told the MES what it was looking at.

Lenspathio is not a replacement for e-beam review or standalone metrology. It is a high-throughput optical inspection and classification layer that integrates into the fab’s existing SECS/GEM infrastructure — designed to be the first filter in the yield decision chain, not a post-hoc analysis tool.

Lenspathio engineering workspace in downtown Denver

Design Principles

Three constraints we design around first

False positive rate is a cost, not a tradeoff

Every false positive generates operator time, tool occupancy, and lot delay. We treat the 0.1% FPR target as a hard engineering constraint — not a marketing claim. The spec is validated on production wafers, not held-out synthetic test images.

Every decision is traceable

Classification output includes defect coordinates, defect type, confidence score, and the model version that produced the call. Lot records are GEM 300 E40-compliant and auditable. No black-box decisions that can’t be reviewed at the time of lot disposition.

Integrate with what fabs already have

We don’t ask process engineers to install new infrastructure, change MES schemas, or manage cloud endpoints. SECS/GEM II and on-premises deployment are the baseline the fab operates from. We start there.

Team

The team

Diana Muller

CEO & Co-Founder

Process integration engineer on 7nm and 5nm logic lines for six years before co-founding Lenspathio. Her evaluation criteria for optical inspection tooling became the product specification for the Lenspathio platform.

Marcus Reyes

Co-Founder & CTO

Computer vision researcher focused on physics-based image synthesis before shifting to semiconductor applications. Designed the CNN architecture and the optical physics augmentation pipeline that generates training data without requiring real labeled wafer lots.

Amara Osei

Head of Fab Integration

Eight years in SECS/GEM equipment engineering across multiple 300mm fab environments. Leads every customer integration engagement from initial E40/E87 protocol setup through production qualification sign-off.

Kenji Tanaka

Process Engineering Lead

Yield analysis engineer with experience across logic foundry and 3D NAND memory lines. Runs customer evaluation engagements and calibrates defect taxonomy parameters to match each customer's specific process step and defect types of interest.

Elena Vasquez

Principal ML Researcher

Research background in physics-informed neural networks and optical simulation. Maintains the synthetic training data generation pipeline and validates false positive rate specifications against real wafer datasets at each supported process node before every model release.

Location

Denver, Colorado

Lenspathio is based in downtown Denver at 1700 Lincoln Street. The team works across Denver and on-site at customer fab locations during evaluation and qualification engagements. We do not have a cleanroom or inspection hardware on-site in Denver — the platform is deployed and validated at customer facilities.

Lenspathio
1700 Lincoln Street, Suite 2000
Denver, CO 80202
[email protected]
+1 (303) 555-0163

Evaluations run on your actual lot data.

If you’re a process or yield engineer assessing inspection options for a sub-10nm fab line, send us a lot and we run the pipeline on your wafer data — not a demo dataset. Results include per-die classification, confidence score distribution, and FPR at your process node.

Request Evaluation Contact us