Every hyperscale buildout in the country is stuck behind the same bottleneck: nobody can get transformers fast enough, and the ones they do get waste energy and floor space the industry can no longer afford. We are a small team replacing that entire box with silicon carbide — and we need people who want their fingerprints on the hardware.
Arguments here get settled with a model, a measurement, or a bench setup. Seniority is not evidence.
Whoever designs the board debugs it at 2am and signs off on the production build. No throwing work over a wall.
You will touch things outside your title constantly. That is the deal, and it is the fastest way to get good.
Hardware in a customer's rack teaches us more than another quarter of simulation ever will.
A design flaw raised in week two is a conversation. The same flaw found in week twenty is a recall.
Medium voltage, SiC, magnetics, controls, thermal — no single engineer covers that span. We teach each other.
Our solid state transformer takes 35kV off the utility and hands 800V DC directly to a rack of GPUs, in one conversion stage instead of five. Getting there means a cascaded H-bridge front end, eighty-four high-frequency transformers switching above 50kHz, a resonant output stage, protection that clears a fault far faster than any mechanical breaker, and liquid cooling running through all of it. Somebody has to own the converter hardware that makes that real, from the first block diagram to the units rolling off a production line. That is this job.
You will sit at the center of a deliberately small hardware team, which means the design decisions you make show up in the product within months, not years — and so do the mistakes. If you have spent the last few years watching your work disappear into a process, this is the opposite of that.
This list is deliberately short. We care far more about first-principles depth, hands-on instinct and initiative than about a checklist. How much domain experience you bring mostly determines the level we hire you at — Senior or Staff — not whether we hire you.
Send a resume and, if you have one, a short note about a converter you took from blank page to working hardware — what broke, and what you did about it. We read those first.
A megawatt-class solid state transformer is a software product wearing a steel jacket. Grid-forming control, protection coordination, the sequencing behind fast fault protection, thermal management, telemetry — all of it is firmware, written by several people, running on hardware sitting between a utility feeder and somebody's production GPU fleet. Getting each piece right is not the hard part. Getting all of it to converge into a release that is safe to energize is.
This role owns that convergence. You are the person who decides what goes into a release and what waits, who knows whether the validation behind a change is actually sufficient, and who says go or no-go. When a unit in the field takes an update, it does so because you were satisfied. There is no separate QA organization to hide behind — the bar is whatever you hold.
This is a judgment role more than a coding role, but not a management one — you will be in the code, on the bench, and in the release meeting the same week. Level (Staff or Senior Staff) follows the depth you bring.
Send a resume plus a short account of a release you held back, and what it cost to make that call. That tells us more than a list of tools.
The two posts above are the load-bearing hires. Everything below is the engineering team that ships the product with them — power stage, controls, magnetics, packaging, mechanical, thermal and EMC. These are hands-on execution roles for engineers who have taken medium-voltage or SiC power hardware into production. Tags show experience level and how urgently we need the seat.
The hands-on board-level engineer who takes a power stage from SPICE model to laid-out PCBA to a working, tested board — and repeats it across the cascade.
Our senior/staff engineers architect the converter; you build the boards that make it real. You take each power-stage block — the cascaded H-bridge cells, the resonant output stage, the LV rails — from a SPICE model to a laid-out PCBA to a debugged, characterized board, then do it again until it's production-ready.
Own the DSP and communications boards end to end — schematic → layout → debug — including board-to-board links, system-level sampling and host communications.
The control brain of the converter runs on the boards you design. In a cascaded system with dozens of cells, the communications fabric — board-to-board, system-level data sampling, and the host/upper-computer link — is as load-bearing as the power stage, and it has to stay clean in a high-dV/dt environment.
Build the hardware-in-the-loop rigs — control-board and full-unit HIL — for a medium-voltage cascaded converter, and stand up the real-time simulation platform behind them.
You can't safely energize a 35kV converter to test every control change, so the HIL platform is how the controls team moves fast without blowing things up. You build that platform from scratch — control-board HIL through full-unit HIL — and keep it faithful enough that a green light on the rig means a green light in the cabinet.
Develop the control algorithms and protection strategies, then carry them from simulation through bench debug on high-frequency, high-power converters.
Grid-forming control, cell balancing across the cascade, resonant-stage regulation and the protection strategy that coordinates with fast fault clearing — all of it is algorithm work that has to survive contact with real hardware. You own that algorithm layer from the whiteboard to the bench.
Design the packaging for our SiC power modules — the layout, interconnect and thermal path that lets the die switch hard and stay reliable.
The SiC die can only switch as fast, as cool and as reliably as its package allows. You own that package — the substrate, interconnect, thermal path and the parasitics that decide whether the module thrives above 50kHz or rings itself apart.
Own magnetic design across the system: board-level auxiliary supplies, transformers, chokes and the medium-voltage-side reactor.
Magnetics is where this converter's density is won or lost. Across board-level auxiliary supplies, the high-frequency transformers, chokes and the MV-side reactor, you do the electromagnetic design that hits the loss, size and temperature targets the rest of the product depends on.
Electromagnetic design and test of the main high-frequency transformer, plus the auxiliary power supply that rides with it.
Eighty-four high-frequency transformers, switching above 50kHz, carry the power across the isolation barrier of this converter. You own the electromagnetic design and test of that main transformer — and the auxiliary power supply that goes with it.
Own transformer insulation end to end — insulation design, the manufacturing process behind it, and the test that proves it holds.
At medium voltage, the transformer lives or dies on its insulation. You own that system — the insulation design, the process that builds it repeatably, and the testing that proves it withstands the field it sees for the life of the product.
Mechanical design for board-level HF transformers, chokes, relays and SiC cooling, and for the PEBB power module — structure, field-strength simulation, insulation and thermal.
The PEBB — the power-electronics building block that repeats across the converter — has to hold high-frequency magnetics, relays and SiC modules in a compact, high-field, liquid-cooled package that survives volume production. You own its mechanical design, from the board-level components up to the module structure.
Full-unit electrical drawings, layout, cabinet and enclosure, mechanical switchgear, installation, and airflow / thermal routing at the container scale.
The finished product is a medium-voltage, liquid-cooled cabinet that ships in a container and gets installed next to a data center. You own the system-level mechanical design that makes that whole box real — layout, enclosure, switchgear, install and thermal routing.
Design the container liquid-cooling system — cooling loop, deionized water, flow and pressure drop — from simulation through implementation and test.
Every watt of loss in this converter ends up in the coolant, so the liquid-cooling system is what lets the box run at the density we're targeting. You own that system end to end — the loop, the deionized-water design, the flow and pressure-drop budget — from simulation to a tested, working system.
EMI design and test — filter design, shielding and grounding — plus running the EMC compliance campaign to a pass.
A megawatt-class converter switching SiC above 50kHz is an EMI generator until someone designs it not to be. You own that — filter design, shielding and grounding strategy from the start, and the EMC test campaign that turns "should pass" into a certificate.
The list above is where we're focused this cycle, but it isn't a fence. If you can make a real dent in a medium-voltage power conversion problem — in manufacturing engineering, firmware, test, or something we haven't thought to post — write to us anyway and tell us what you'd want to work on: careers@lyrahtech.com.