Case studies · Foxit Software, 2020–2026

Five platforms, told as short stories.

For each: the situation, the problem, what I decided, what I did, and what happened. Plain words where plain words will do.

2020

From a two-day wait to fifteen seconds

PDF Developer API SaaS platform

Situation: The pandemic had just started. A developer needed about two days to get a trial key for Foxit's PDF SDK, which was sold as a download.
Problem: There was no self-service cloud product, just as customers moved to remote work.
Decision: Prove it with working software, not slides. I pitched a prototype to top management, then built it as the only application engineer in the unit (the PDF engine stayed with R&D).
Action: I coded the whole product in C#: sign-up, instant API keys, documentation, SDK samples and a live API console. First iteration in two weeks.
Result: Onboarding went from about two days to about fifteen seconds. It reached production, and the founder asked every team to use it as the reference for a real SaaS product.

2021

End-to-end e-signature product

Situation: Foxit wanted a backup e-signature product in case an acquisition fell through. Four engineers in China were assigned to me; I had been in R&D for three months.
Problem: Build a complete product quickly, with a new team, while keeping management confident it would work.
Decision: One-week sprints that always end in working software, and an open 30-minute review that anyone in the company could attend.
Action: I acted as Product Owner and Scrum Master, ran every review and recorded the demos.
Result: A working end-to-end signing flow in week one and a complete first version at sprint 12. The founder called the demo "very good" and the Japan business took it up. The acquisition then succeeded, so the team moved on.

2022–23

Microservices turnaround for a 15-person team

Situation: A new construction-industry PDF product, with about 15 people in Germany and Ireland. The engineering VP and the Product Manager were not doing their Scrum roles.
Problem: There was no working software and no shared architecture.
Decision: Show by doing: cover the Scrum Master and Product Owner roles myself for two sprints, and build a reference architecture the team could copy.
Action: I wrote a multi-service reference implementation on .NET and Dapr (service calls, messaging, state, tracing, logging, metrics). From sprint three I handed the roles back and coached the VP and Product Manager.
Result: An MVP in sprint one and improvements in sprint two. The team said they hadn't believed they could do it. The architecture was fully adopted within about six months.

2023–26

Enterprise data platform

Situation: I had no assignment. Finance had just hired its first data analyst, and few people knew what a data platform was.
Problem: Technology alone would fail. Each department had to own its data and want to use it.
Decision: Build the platform and the habit together.
Action: I built it hands-on on Azure (Data Factory, Databricks, Purview, Power BI). I wrote a data team handbook with request intake and prioritisation, asked each department head to sponsor a team, and ran an internal hub explaining it.
Result: Product, Finance, Sales, Marketing, Operations, IT and R&D use it, and key departments run their own data teams on it.

2025–26

Foxit's internal AI platform (FIAP)

Situation: Staff were trying AI tools on their own, which put company data at risk.
Problem: Make AI useful in daily work without leaking data.
Decision: Build shared foundations and a governed gateway, and ship ready-made tools people could use straight away.
Action: I coded it (C#, Windows and macOS): connectors to Microsoft 365, Salesforce and Jira, a knowledge graph with short-, mid- and long-term memory, and a gateway that sent all AI traffic, including coding tools, through approved providers with PII redaction and usage tracking.
Result: Daily use in Sales, Marketing and R&D. One user said the meeting-notes tool beat a commercial product for their language; another: "I use your tool every day."