The problem
In a high-volume contract-manufacturing environment, the product changes but the discipline doesn't: every process has to be fast, in control, and documented well enough that quality doesn't depend on any one shift. Optical assembly adds precision and cleanliness constraints on top of standard SMT, so there's little room for a process that drifts.
I did two tours at Fabrinet, and they turned out to be the two ends of the same pipeline: process engineering on NPI products and incoming technology transfers in Thailand first, then running the advanced packaging lab in California — developing the processes that went the other way, out to mass production in Thailand. Writing a transfer package for a factory you have already worked inside is an unfair advantage, and I had it.
Santa Clara, California · 2017 – 2019 · internship 2016 – 17Advanced Optical Packaging Engineer
Running the Advanced Packaging Lab
Ran the Advanced Packaging Lab — the room where die attach, wire bonding, and optical attach processes were developed before they were allowed to become production processes. Developing them was only half the job: each one then had to transfer to mass production in Thailand and survive there, on other people's equipment and other people's shifts, which is a much harder test than working in the lab it was born in. Multiple programs for multiple customers went down that path — NPI in the lab in Santa Clara, then transfer to production in Thailand. In production, those products have since passed $200M in revenue.
Optical packaging process development
Owned process development for new products across the whole packaging flow — die attach, wire bonding, encapsulation, and plasma cleaning — for several customers and projects running in parallel. On die attach that meant developing adhesive, eutectic, and solder-preform processes across chip-on-board, chip-on-carrier, and chip-on-wafer, and owning the tools as well as the recipes: better MTBA and MTBF on the bonders was part of the job, not somebody else's.
SMT & precision assembly process
Owned and optimized process steps across print, place, reflow, and precision optical assembly — profiling, setting process windows, and qualifying changes before they reached the line.
Designed experiments, not opinions
A persistent delamination problem in lens and mirror attach came apart under a proper DOE — temperature cycling across the variables, then detailed metrology on the results — rather than under argument. The same rigor went into 8D and MRB reporting: every investigation ended in a recommendation somebody could act on, not a description of what happened.
The first version of the data system
Built a wafer-assembly tracking system in VBA and deployed it on the Amicra die bonders: it recorded which diode went to which wafer location, reported yield and dropped-diode counts at the end of each wafer, and put a wafer on engineering hold automatically if yield fell below 85% instead of letting it move to the next operation. Nine years before the MES, the instinct was already the same one — the process should report its own state, and the limit should stop the line, not a person noticing.
Process Engineer II
NPI bills of material & assembly procedures
Examined and refined the bill of material and assembly procedures for products coming through NPI, and wrote the assembly work instructions — the MPIs — that the floor actually built to. A good chunk of the job was factory efficiency and task prioritisation, worked out by studying how product physically moved through the plant, which is the kind of analysis that only works if you walk the line.
Debug, root cause, and $220k of MRB
Tightened the debugging and root-cause process on recurring defects, and the MRB bill came down by $220,000. Cost that size doesn't come out of one heroic fix — it comes from making the path from "this failed" to "this is why" short enough that material stops sitting in review.
International technology transfer
Led the transfer of two international fiber-rod technology projects into the plant, one from France and one from Singapore — taking a process that worked somewhere else and making it work here, with the documentation and training to keep it working after the visiting engineers flew home.
Training the floor
Trained and mentored new technicians and operators. It belongs on this list: a process is only as capable as the people running it at 2 a.m., and this was the start of the part of the job I now do full-time.
The takeaway
Volume rewards rigor. The gains came from treating each process step as something measurable and improvable — and from building the controls that keep a tuned process tuned after the project closes. Six years apart, on two continents, the method didn't change.