Alexander
Vervloet.
For eight years I shipped production web and mobile software, six of them on a platform with two million users, owning flows end to end from React through GraphQL on NestJS to Kafka. These days the systems I build are agentic ones: tool-using agents allowed to do irreversible things, and the durable execution, approval gates and eval suites that keep that safe. Same engineering, harder failure modes. Alongside the building I kept doing the product work: customer surveys, scope arguments, prioritization, and explaining system behavior to executives who had never written code. That half of it is further down this page.
The engineer who kept asking product questions.
At VeVe I was never only the person who built things. I was the person who asked why we were building them, whether we were building the right thing, and how to explain the answer to someone who had never written a line of code.
Over six years that stopped being a personality trait and became a standing part of the job. Surveys to find out what users actually wanted. Prioritization sessions where I argued the queue. Scope conversations with leadership, held in their terms rather than mine. PRDs I helped write because the spec needed a technical reality check before anyone committed to a date.
I still want to be the person writing the code. But I'm at my most useful when I also have a say in what gets written, and I read a two-week estimate against a two-month one before anyone has opened an editor. Teams that want both from one hire are the ones I'm looking for.
- →Asking why, not just how.
- →Turning an architecture argument into a business one.
- →Feedback in planning that changes the plan.
- →Reading a spec and seeing the four weeks nobody costed.
- →Knowing when not to ship.
- →Teaching — four years, in two languages.
Eight years. One long streak.
- 01Building agentic systems from scratch, without frameworks, so the primitives are the thing I understand rather than the libraries wrapping them. Tool-using agents, durable execution, approval gates, retrieval pipelines, eval harnesses, and prompt-injection defenses, each a runnable project with measured results.
- 02Shipped four distinct flagship projects that span building, running, and orchestrating LLM systems: askrepo, a codebase-Q&A app answering with (path:line) citations across eight eval-gated stages; a self-hosted local-LLM news pipeline with its own eval suite and guardrails; a harness that stress-tests the spec-driven-development workflow with gated phases and deterministic traceability checks; and Knowledge Desk, a deployed multi-tenant assistant whose permission boundary is enforced three independent times.
- 03Publishing the work as an open, teachable series spanning RAG, agents, evals, guardrails, MCP, multimodal, fine-tuning, and local models.
- 01Owned the web payment flow for in-app currency end to end, from React frontend through the GraphQL API on NestJS to Kafka event publishing. Took web from no purchase path to 20% of purchases. Millions of dollars processed. Zero critical payment bugs shipped.
- 02Built the mobile purchase UI up to the App Store and Google Play handoff, as the engineer mainly responsible for it. That flow carried an estimated $112.5M in consumer spending by late 2021, per Sensor Tower.
- 03Led the migration of platform search from Redis to Elasticsearch on Node.js and NestJS, raising measured search satisfaction from 41% to 90%, then built a carousel backend service on the same infrastructure.
- 04Led the rebuild of the web platform from a client-side React app to Next.js across 15 frontend and backend engineers. Server-side rendering and caching cut average page load from 5 seconds to under 1 second. Ran the education session for engineers, PMs and leadership before kickoff, and the deck became the onboarding reference.
- 05Led the Collectibles team to move part of the RabbitMQ flow to Kafka and build drop subscriptions end to end. Buyers are drawn from a subscriber pool and unclaimed purchase rights are reassigned after a timeout. Used on thousands of drops since, with about 50,000 subscribers on the most popular.
- 06Built the web storefront, auction bidding system, and direct-purchase flows from scratch, against the GraphQL API that every client on the platform read from. These became the primary revenue surface for hundreds of thousands of transactions.
- 07Built the 10-step mobile onboarding flow for the app launch, which supported tens of thousands of signups in the first weeks. Built the guided tour that closes it and tuned it against feedback and usage data, raising repeat logins after signup from 8% to 24%. Also built the web wallet frontend with ethers.js and web3.js.
- 08Ran feature-specific customer surveys with the Lead Data Engineer, shipped the recommendations behind feature flags for A/B testing, then re-surveyed to check whether they had landed. Overall customer satisfaction rose nine percentage points.
- 09Helped lead feature and bug-fix prioritization inside a domain team for three years, frequently as Scrum Master, rebalancing the queue each week against results, customer needs, and company objectives.
- 10Won leadership support for folding Crafting into the existing store rather than building it as a separate section. Compared the two workflows, presented the delivery and maintenance tradeoffs, and saved an estimated three to four weeks of frontend work. That slack is what absorbed the late scope changes without moving the launch date.
- 11Contributed feature context and technical specs to PRDs, and argued for QA checkpoints during development instead of at the end by putting the risk in delivery-date terms. That feature shipped early with fewer bugs than anything else we shipped that quarter.
- 12Built out Amplitude event tracking and Datadog observability across the apps, so questions about product behavior had data behind them instead of opinions.
- 13Served as the technical translator between engineering and product, brought into executive meetings specifically to explain complex system behavior in plain terms. Most active contributor in every planning session across six years, and regularly the only person asking whether we should build the thing at all.
- 01Led frontend in a small agile team building a React platform matching brands with influencers — shipping weekly through constantly changing specs.
- 02Grew customers reporting satisfied or very satisfied from 200 to 1,000 through targeted testing, tooling improvements, and systematic refactoring.
- 01Built and maintained frontend features using React, React Native, and D3.js in a fully remote agile team, coordinating across time zones from day one.
Decisions, not just delivery.
Shipping was never the whole job. These are five calls I made at VeVe, each with the situation, the reasoning, and what actually happened, including the one that ended badly. Nobody assigned me most of this work. It was the part I kept reaching for, and it is the part I want more of.
The spec called for Crafting to be its own section of the app. Mapped against the store we'd already shipped, the overlap was almost total, and the genuinely new surface came down to two things. I took the architectural argument to leadership as a business one.
Shipped ahead of schedule. Three to four weeks of frontend work saved, and the slack is what absorbed late scope changes without moving the launch date.
Leadership assumed the team was ready for the App Router. One-on-one, most engineers admitted they weren't, and we were days from kickoff. Nobody asked me to fix it. I taught the whole team, PMs and leadership included, because the architecture had product consequences they'd need for scoping.
Building productively within days instead of weeks. PMs used the terminology correctly in tickets from the start. The deck became the onboarding reference.
You could predict how well a team delivered from its communication habits. Nobody owned the problem. I watched quietly for weeks first, then ran working groups picked for willingness to be critical, so the teams felt they'd written it themselves.
Became the de facto standard, part of new-hire onboarding, and cited in scope arguments. Still in active use two years after I left.
Most wallets are built by crypto people for crypto people. Our users arrived for the collectibles and ended up holding a token, so I couldn't assume they knew what a pending state, a rejection or a chain mismatch was. The price of misreading a screen was their own money, permanently.
Still in production four years later, with no engineer pulled back to it. Getting the failure states right up front is what bought the silence.
I found out my role was on a layoff list that was already final. The letter had to do two jobs: argue on the merits that the proposed engineering cuts went further than the strategy required, and make an evidence-based case for six years of my own contributions. Either one alone fails.
Leadership reversed the decision. A later round in November 2025 took the role anyway, which is the honest ending.
"His instincts around product delivery and team health would make him a strong asset in any Product Manager role."
"He always took the time to explain technical things in business language, so everyone could understand and contribute better to discussions."
The stack, and the product work around it.
- TypeScript · React · React Native
- Next.js · Vite · Tailwind
- Node.js · NestJS
- GraphQL · REST · SSE
- PostgreSQL · Elasticsearch · Kafka
- Payments · Web3
- Docker · AWS · CI · testing
- Agents · tool use · MCP
- Durable execution · approval gates
- RAG · vector search
- Evals · LLM-as-judge
- Prompt engineering · guardrails
- OpenAI · Anthropic · local models
- Python · FastAPI
- Customer research · surveys
- Prioritization · scope tradeoffs
- PRD contributions · requirements
- A/B tests · feature flags
- Scrum facilitation
- Amplitude · Datadog · SQL
- Technical-to-business translation
- Cross-team collaboration
- Stakeholder communication
- Developer training
- Explaining complexity simply
Things I build outside the job.
The series I wrote to learn it.
Building in the open.
Four years in a classroom.
Before moving into full-time engineering I taught English and core subjects at schools in Taiwan — including one of the country's top-ranked high schools. Math, science, social studies. Often in Mandarin. Gamification and question-driven methods.
I also tutored privately — students from age 6 to working professionals, on everything from English to beginner Python and Raspberry Pi hardware.
The discipline of making complex things genuinely simple for people with no prior context has turned out to be the single most useful thing I do.
- English
- Native
- 中文 Mandarin
- Intermediate · self-taught, 12 years in TW
- B.S.
- Communication — Speech, Rhetoric, Leadership
- School
- Oregon State University, 2014
- Major GPA
- 3.65
Start a conversation.
I'm looking for remote software engineering roles, ideally ones with agentic or applied-AI work in them, on teams that treat product thinking as part of the job rather than someone else's department. I'm equally open to product or technical PM roles, where eight years of shipping the thing is the background I'd bring to deciding what to ship. UTC+8, flexible overlap with US and EU.
- alex.vervloet [at] gmail [dot] com
- alexander-vervloet
- GitHub
- @alexvervloet