You were probably looking for slowey.dev — the AI coworker that lives in Slack. This slowey is on-prem in HCMC, deployed at Katalon, and runs on cà phê sữa đá. Too good to pass up.
Three live surfaces: request threads, toolchain context-switching, and incident memory. Click anything.
Requests come in noisy, outcomes go out shippable. Usually before the second coffee.
@slowey our pipeline is back at 38m. can we split it today?
and cache warmup keeps flaking on retries.
on it. stage split + deterministic cache key. PR before standup ends.
Context-switches on demand. Click a tile to inspect the active subsystem.
routing + eval loops
Production recall works like a timeline. Pick any entry to load the remembered response pattern.
Matched a previous cache-stampede signature from Q1.
Added singleflight guard + warm cache path before traffic cutover.
Twelve projects across full-stack, cloud-native, blockchain, ML, and AI Product. Six featured below; the rest behind a disclosure.
Web dashboard for the llmfit OSS project. Hardware-aware LLM model recommendation, filtering, side-by-side comparison.
24/7 AI support chat across the Katalon suite. LangGraph + Claude 4.5 Haiku with web search; Salesforce handoff for Enterprise.
Microfrontend + .NET 8 microservices on AWS EKS with Istio. Module Federation host shell + CQRS services + full observability.
Carbon credit + footprint platform. Six Go microservices, Kafka event bus, Traefik gateway, blockchain-verified offset certificates.
Fan-engagement web app where T1 esports fans light virtual incense for tournament victories. Tournament-timed launch.
Blockchain e-notarization. Solana NFT certificates, AI document verification (Gemini), real-time signing sessions, PayOS payments.
Contributor-friendly ML experiment manager. PyTorch CNN/MLP/RNN on MNIST; WebSocket-streamed loss + accuracy.
Full-stack resort ops: reservations, self check-in/out, housekeeping, PayOS billing, multi-role analytics.
Full-stack contest hosting: organizer flows, registration, PayOS payments, Cloudinary media, Firebase notifications.
React Native shop ops + Node/PayOS gateway. UIT SE104 team project, 5 contributors.
React Native t-shirt + tote-bag designer with marketplace listing, internal currency for profit margins, PayOS checkout.
Traditional ML emotion classifier (SVM + Random Forest on HOG features). React Native + Flask. Companion ResearchGate paper.
Three steps. Each one slower than installing a Slack app, but more durable. No tokens to rotate. No cloud computer to provision. No recruiter pitch necessary — he's already deployed.
Open resume.slowey.dev. Two pages. Takes 4 minutes. No commitment, no NDA, no recruiter pipeline.
To hi@slowey.dev. Memes, technical questions, OSS pings, conference hellos — all welcome. Recruiter pitches go straight to /dev/null.
HCMC preferred — cà phê sữa đá at any quận 1 spot. Async equally fine. "I read your blog" through "explain your LangGraph migration" all welcome. "Wanna leave Katalon?" does not — he likes it there.
Stable API surface across the verticals you actually need. No fine-tuning required (pre-fine-tuned by 4 years of full-stack engineering).
Roadmap, stakeholder alignment, customer feedback loops, Jira/Agile. Translates technical complexity into business value.
Java / .NET / Go / Python · microservices · gRPC · CQRS · event-driven. Clean architecture per service, not per repo.
AWS / Azure · Kubernetes · Helm · Istio · Terraform · CI/CD · observability stack (Prometheus / Grafana / Jaeger).
LangGraph · LangChain · FastAPI · PyTorch · RAG · Claude integration. Built the Katalon Support Assistant on this stack.
Test pyramids that aren't sand pyramids. axe-core, Lighthouse, Playwright. Documentation that future-you can grep.
Field-tested in HCMC cafés. Maintains a 99.4% delivery SLA on technical reasoning under cà phê sữa đá load.
Tested against frontier models on the workloads that bill at the end of the month: deflecting tickets, surviving on-call, ordering coffee.
Methodology: Katalon Support Assistant case-volume reduction over 8 months in production. Other models declined to integrate with Salesforce.
Methodology: PRs merged into AlexsJones/llmfit between Feb–Apr 2026. Dependabot is admittedly very competent at bumping versions.
“Cho tôi xin một ly cà phê sữa đá” delivered with conviction. Improving steadily.
Tool-use built in. No MCP server required (he is the MCP server).
Type-safe wherever possible, pragmatic everywhere else.
No GPUs required. Sleep allowance is non-negotiable. Coffee is the rate-limiting reagent.
Click any role to load the changelog. No rollbacks recorded.
Real teams shipping real features with slowey (the human, not the chatbot at assistant.katalon.com).
“Came for the OSS PR, stayed for the cà phê sữa đá memes. The dashboard ships.”
“Said 'I'll have a look' on Slack at 09:14, fixed it by 09:31. Still not sure how. Five stars.”
“Tried to file a support ticket on assistant.katalon.com, accidentally bookmarked the human. Now check his GitHub weekly. Five stars, would confuse again.”
Stable API across the channels your org already uses. Webhooks not required.
No tokens, no rate limits, no per-seat surprises. One slowey per organization (singleton by design) — and his current org is Katalon.
Most of these come up. The rest came up once and seemed worth keeping.
He's already running in production at Katalon. Read the CV out of curiosity, or just say hi — coffee, memes, and Terraform questions all welcome.
