On site, in the codebase, on the schedule.
David T Phung is a Forward Deployed Engineer for the parts of the economy that still run on paper. Eight years embedded with owners, trades, agencies, and software teams, $125M+ in delivered portfolios, and a 1M-user platform shipped on the software side. He is what most labs are actually trying to hire when they post the FDE role: a domain operator who has also run a release train.
A frontier lab’s FDE has to do four things at once. Speak the customer’s domain. Write code that ships next week. Coordinate the customer’s stakeholders. Carry the outcome back to product and engineering.
Most candidates can do two of those. My track record is in all four, and the four moves below are the evidence: an RFI triage system, a $200 energy audit, a 12-API consolidation, and a parallelized permitting cycle. Each one is the same shape as the work a Forward Deployed Engineer ships.
- $125M+ delivered across luxury retail TIs, civic, and F&B.
- 1M-user platform · 99.98% uptime · 10 engineers, 2 designers.
- Production facility built from zero, scaled 460% in two years.
- RFI triage at 1.8d vs the 7 to 14d industry baseline.
- 12 APIs consolidated; reconciliation time cut 60%.
- Three agencies cleared in one review cycle, 7 weeks saved.
- B.Arch, Woodbury University, minor in Urban Studies.
An RFI triage system, on a $25M+ portfolio.
At Gruen Associates, four simultaneous retail and civic projects, 20+ consultants, and four general contractors. The RFI queue was the bottleneck. I designed and ran the triage system that compressed it.
Of $340K in change orders evaluated, I rejected $155K with documented justification. Holding the CO rate at 3% against an industry baseline of 8 to 12% is the kind of margin defense an FDE shipping into procurement or claims teams has to produce.
A $200 device, a 65% energy cut.
At Happy Friday Coffee, I ran a hardware-instrumented audit on a production facility I had built from the ground up. The cheapest sensor I could find, pointed at the right meter.
Saw a question existing telemetry could not answer, bought the cheapest device that could, ran the experiment, found 40% waste during non-production hours, and shipped the SOP that captured the savings. Hardware sensor, software methodology.
12 APIs to 1 surface, 60% reconciliation cut.
Product owner at Entertainment Partners on a 1M-user platform across the US, Canada, and the UK. The reconciliation cost between 12 integrations was the dominant operational tax. I led the release train that collapsed it.
Twelve heterogeneous APIs that finance, ops, and engineering each consumed differently, collapsed onto one unified surface. The deliverable was not a new platform. It was an alignment artifact: one source, one contract, one reconciliation flow.
Three agencies, one review cycle.
Permitting is where physical-world software projects die. On a Gruen Associates job, three agencies had to clear before the schedule could move. I ran them in parallel instead of in series.
Coordinated city, county, and federal reviews on one project, prepared a single packet that satisfied all three intake teams, and got the approvals back on a parallel rather than sequential schedule. The skill is not lobbying. It is reading how each agency's intake actually works, then sequencing so the dependencies collapse.
What I bring that a pure SWE and a pure domain hire each cannot.
Domain literacy that you cannot fake.
- B.Arch from Woodbury. I read a drawing set and I can tell you what the field will actually do with it.
- Construction administration depth: RFIs, submittals, pay applications, change orders, commissioning, turnover. Each one is a workflow a software team would otherwise have to discover from the user the hard way.
- Trade fluency: Type I/II hoods, Ansul, grease interceptors, makeup-air balancing, walk-ins, heavy-power electrical. I have specified and installed every one of those.
- Eight years sitting between owners, designers, trades, agencies, and software vendors. The integration is already done in my head.
A release train, at scale.
- Product Owner on a 1M-user platform at 99.98% uptime, across the US, Canada, and the UK.
- Led a 10-engineer, 2-designer release train aligned with finance and operations on one schedule.
- 12 APIs consolidated into one surface. 60% reconciliation time cut. The integration muscle a lab’s product surface needs.
- Tools at work: Jira, Confluence, Procore, Bluebeam, AutoCAD, GIS, Excel, MS Project. Fluent in both stacks.
Where this hybrid actually pays back.
Forward Deployed Engineer at a frontier lab
Embedding with industrial, AEC, or regulated customers. Translating a customer’s actual workflow into shippable software, instrumenting outcomes, then carrying the lesson back to product and engineering.
Founding Solutions / GTM Engineer at a vertical AI startup
Construction, real estate, infrastructure, energy, F&B operations. Speaks the trade, owns the integration, and builds the first ten reference customers so the product can find its shape.
Field CTO / Customer Engineering Lead
PropTech, ConTech, AEC software, energy software. Owns the seam between the customer’s field reality and the platform team. Already worked it from both directions.
Forward Deployed Operator / Owner’s Rep, software-native
Real estate and infrastructure development that runs on software. The traditional owner’s rep, with a release train’s rigor and a product owner’s instrumentation reflex.
Builder to builder, thirty minutes. If the role is right, we will both know inside the half hour.