We’re pleased to feature yieldWerx, predictive semiconductor manufacturing analytics and automation software developer, in our latest MicroLED Industry Association spotlight series.
We spoke with Aftkhar Aslam, yieldWerx's CEO and co-founder, to learn more about the company's technology, business, achievements and goals.
Can you introduce your company and technology?
yieldWerx builds yield management and test data analytics software for the semiconductor industry. Our platform pulls together data from across the manufacturing flow (wafer sort, final test, inspection, assembly) and turns it into something engineers can actually work with: wafer maps, bin and parametric analysis, defect pattern detection, and automated reporting. We describe it as yield intelligence throughout the digital thread, from design all the way to finished goods.
We've spent close to two decades doing this for traditional semiconductors, working with IDMs, fabless companies, and OSATs around the world. What's changed recently is the scale problem. Emerging technologies like microLED push die counts and data volumes far beyond what conventional yield tools were built for, and that's exactly the problem we've been engineering toward.
Why did you join the MicroLED Association, and what do you hope to achieve?
MicroLED is fundamentally a yield problem. Everyone in this industry knows the math: a single 4K display needs roughly 25 million individual microLEDs, and even 99.99% yield leaves you with thousands of defective pixels. Getting from impressive demos to affordable volume production will be won or lost on yield learning: finding defects faster, understanding their root causes, and making known-good-die decisions before mass transfer rather than after.
We joined the Association because we want to be part of solving that alongside the people building the epitaxy, inspection, transfer, and repair technologies. Our goal is practical. We want to work with members to understand where their data bottlenecks are, adapt our analytics to microLED-specific workflows, and help establish common ground on how yield and test data should flow through a supply chain that's still defining its standards.
What is your biggest challenge, and success to date in the microLED industry?
The biggest challenge is scale. A conventional IC wafer might carry a few thousand die; a microLED wafer carries millions of emitters, each of which needs to be characterized, mapped, and dispositioned. Most yield systems simply fall over at that volume. The wafer maps won't render, the queries take hours, and engineers end up sampling instead of seeing the full picture. Sampling is dangerous in microLED, because the defects that kill you are often subtle spatial patterns that only show up when you can see everything.
Our success has been proving that full-resolution analysis at that scale is workable. We've demonstrated our platform handling more than 5 billion data points per wafer interactively — pan, zoom, filter, and run defect pattern analysis on the complete population, not a sample — with AI-based pattern recognition flagging spatial signatures like rings, clusters, and scratches that point straight to process root causes. That capability was built for exactly the die densities microLED production demands.
Can you detail your latest prototype or demonstration?
Our most recent live implementation is an end-to-end defect and yield analysis flow built around very high die-count wafers. It ingests inspection and test data, renders complete million-die wafer maps in the browser, and layers analytics on top: defect source analysis across process steps (new versus carried-over defects), killer defect ratio calculations that tie inspection defects to actual pixel and/or electrical failures, cross-wafer commonality analysis to catch repeating and systematic defects, and AI pattern detection trained to recognize yield-limiting spatial signatures automatically.
For microLED specifically, the interesting part is the known-good-die output. The same analysis that explains where yield is being lost also produces the die-level disposition data that mass transfer equipment needs: which emitters to pick, which to skip, and where repair is economical. We're keen to demo this to Association members and tune it to their workflows.
How do you see microLEDs changing the display industry in the next 5–10 years?
We think microLED follows the classic semiconductor trajectory: it starts where its advantages justify the cost (luxury TVs, AR headsets, automotive, wearables) and then works its way down the cost curve into the mainstream. The physics case is settled; nobody disputes that microLED wins on brightness, efficiency, lifetime, and form factor. The next five years are about manufacturing economics, and the ramp will be paced almost entirely by how fast the industry climbs the yield learning curve.
That's the exciting part from where we sit. The companies that win in microLED won't necessarily be the ones with the flashiest demos. They'll be the ones that instrument their lines properly, learn from every wafer, and drive defectivity down faster than their competitors. Displays are becoming semiconductor products, and that means semiconductor-grade yield discipline. In ten years we expect microLED to be the default technology at the premium end of the market and pushing hard into the middle, with the industry's data infrastructure matured to match. The wildcard is beyond displays: microLED's role in AI data centers and co-packaged optics is still taking shape, and it could open a second market that's every bit as demanding on yield.