關於
企業家，軟體工程師和創始人。對人類與技術的交集感興趣。相信增加工程實踐的透明度。利用在日本和台灣的海外生活中的經驗來創建富有同情心和文化敏感性的設計。專注於跨文化設計。專業：Rea…
服務
活動
432 人關注
-
史家瑞 已轉發史家瑞 已轉發Building open source is one interaction at a time on GitHub. Max evangelized Airflow to 50 companies before momentum hit. He was terrified of public speaking but forced himself to give talks. He jumped in his car and drove to LinkedIn, Facebook, small startups, and big companies to show what Airflow could do. "I often say that building open source is like one interaction at a time on GitHub, but like good docs and being present, you know, generally and, you know, giving talks." If you are maintaining an open source project or thinking about starting one, this is the work. Not just the code. The interactions. The trust building. The evangelism. #OpenSource #DeveloperRelations #CommunityBuilding #Airflow #TechLeadership
-
史家瑞 已轉發史家瑞 已轉發As a data team lead, you want one thing: critical validations happen automatically, no matter how experienced your team is. This is your safety net that prevents putting out fires in production 🔥 See how Recce creates this safety net in the blog. Link in comments #dbt #DataEngineering #Analytics #DataValidation
-
史家瑞 已轉發史家瑞 已轉發Most companies treat real-time streaming like dark magic reserved for the top 1% of engineers. On the latest episode of the Data Renegades Podcast, Micah Wylde sits down with CL Kao & I to discuss how this is a failure of tooling, not talent. At Lyft, Micah managed Flink at massive scale. Tens of millions of events per second with an elite platform team managing SQL interfaces, Python wrappers, and custom tooling layered on top. Despite this infrastructure investment, maybe 3 engineers out of 100 could actually build a pipeline without help. Everyone else fell past the SQL layer into Java stack traces and state management hell. That frustration led him to build Arroyo from first principles. Then Cloudflare acquired it. CL and I sat down with him to dismantle the gatekeeping around streaming. We covered: ⭐ Why SQL must be the actual language of the engine, not a thin compilation layer that drops you into complexity when it breaks. ⭐ The broadcast state bug that caused exponential state growth and took down every Flink pipeline at Splunk. ⭐ Why Debezium brings an entire painful Java ecosystem to what should be a lightweight problem. ⭐ Why streaming should not require a PhD in state recovery. Link in comments, or listen to wherever you get your favorite podcasts.
-
史家瑞 已轉發史家瑞 已轉發🚨 NEW PODCAST: In this debut episode of the Data Renegades Podcast, CL Kao and Dori Wilson of Recce dissect the shift from software 1.0 to the emerging landscape of software 3.0. Tune in for this thoughtful exploration of how engineering, data, and community-driven innovation converge in today’s rapidly evolving world. 🎧 https://hubs.ly/Q03VNCx90
-
史家瑞 已轉發史家瑞 已轉發I posted a photo on Friday saying I was "pumped" about this launch. Now that it’s actually here, I’m just really proud. The Data Renegades Podcast is live. When CL Kao and I sat down to plan this, we had one rule: Make it feel like a coffee chat. If you work in tech, you know exactly what that means. It’s the meeting you set up when you want to pick someone's brain, get advice, or learn something new. We launched with three episodes today, and I'm still mulling over these conversations weeks later. We spoke about many topics, but to pull some juicier tidbits: 🎙️ Ep 2: Data Journalism Unleashed with Simon Willison (Co-Creator of Django) is a true believer in democratizing data work. He challenged the industry's disdain for "Vibe Coding," pointing out that the global economy runs on unversioned Excel spreadsheets—and that empowering people to just use data is more important than a perfect stack. 🎙️ Ep 3: Building Tools that Shape Data with Maxime Beauchemin (Founder of Apache Airflow & Preset) discusses why breadth is now worth more than specialization. The era of the siloed specialist is ending. (Big up to us liberal arts degree holders). And for our very first episode, I interview my co-host/boss CL Kao. He shares his experience building pro-democracy open source software. It's a fascinating look at the intersection of code and civic duty. I know your feed is full of noise, but if you want a genuine look under the hood of the data industry (perfect for a holiday travel binge), give it a listen. Thank you to everyone who liked the post on Friday. Expect a DM soon! Links below. 👇 #DataRenegades #Podcast #DataEngineering #HolidayTravel
-
史家瑞 已分享👀 Now this is interesting!史家瑞 已分享🚨 NEW PODCAST: In this debut episode of the Data Renegades Podcast, CL Kao and Dori Wilson of Recce dissect the shift from software 1.0 to the emerging landscape of software 3.0. Tune in for this thoughtful exploration of how engineering, data, and community-driven innovation converge in today’s rapidly evolving world. 🎧 https://hubs.ly/Q03VNCx90
-
史家瑞 已分享Alright folks, if you are spending any time commuting or working out in the gym, I recommend you put on one of these podcasts. What a great way to receive insight into the current way data is being handled, and the future direction the marketplace is going.史家瑞 已分享🎙️ Data Renegades Podcast drops today! Dori Wilson and I are launching with our first 3 episodes featuring the pioneers who shaped modern data infrastructure and continue exploring its frontier. Listen in and let us know your feedback and guest suggestions! 👇 https://datarenegades.com/
-
史家瑞 已轉發史家瑞 已轉發上個週末由 Kent Huang, Jared Scott, Wei-Chun Chang 代表 Recce 參加 PyCon TW 2025 遇到很多熱情的朋友 參展主要目的是招募 Sr. Software Engineer, 從剛畢業的學生到資深工程師都有來攤位跟我們聊聊，甚至有位講者來詢問這個職缺。我們仍在持續招募，已經有不錯的人選陸續面試中，如果你想加入 Recce 一起協助 data teams 請盡快投履歷，連結在留言中。 令人驚喜的是，雖然攤位上寫著大大的 Join Us 以招募為主，仍有許多人跑來想了解 Recee 在解決什麼問題。從台灣到國外都有資料團隊對我們的產品有興趣，尤其聽到我們是 open source 甚至有企業想要直接 donate！ 從這些交流中，我們再次確認 data validation 資料驗證的需求是確實存在的，需要更好的工具解決這個問題。如果你正在設法提高資料品質，希望能在資料改動上線前了解、確保對下游報表的影響，現在就試試 Recce : https://cloud.reccehq.com 感謝有來攤位找我們的各位，每個人的交流、提問或分享都對我們甚有啟發，感謝 PyCon Taiwan We, Recce, had an amazing time at #pycontw last weekend! Our team of 3 engineers (Kent Huang, Jared Scott, Wei-Chun Chang) represented Recce at the booth, and what an experience it was! 🙌 Hiring Update: We met fantastic candidates ranging from talented new graduates to experienced engineers showing high interest. Even had a conference speaker approach us after loving our product and asking about the position! Our role is still open as we continue conversations with promising candidates, so APPLY NOW before we close it! (application link in comment) While we came primarily for hiring, we were blown away by the genuine excitement for Recce itself. 🤩 We had data teams from companies in Taiwan and abroad approach us with real enthusiasm about our product. The magic moment? As soon as we mentioned we're open source, their interest skyrocketed! 🚀 One developer from a large organization even asked if they can donate. Clearly, there’s real demand for better data validation. For those dealing with data quality issues or worried about the impact of changes. Try Recce today: https://cloud.reccehq.com Thanks to everyone who stopped by! The mix of hiring conversations and genuine product interest made this conference incredibly valuable.
-
史家瑞 已發表感受史家瑞 已發表感受I joined an 8-person startup with no GTM function. 12 months later, I'm leaving behind 11 systems, a 200+ person community, & a content engine built on AI workflows I designed from scratch. Today is my last day at Recce. I met CL Kao the week after my own startup journey ended. I was planning to take multiple months time off, but the role was too interesting to pass up. It was a mix of the growth work I'd been doing at my own company & getting back to the end-to-end data work I enjoy. CL pitched me a role that was broad by design, and broad was an understatement. Being the first GTM hire at a pre-seed company meant I built whatever needed building. A podcast. A meetup. A partnership program. A deanonymization-to-engagement pipeline. An AI-optimized content layer for LLMs. An automated changelog. Organizing an Unconference. Taking sales calls. Building out the internal repo to best practices for better dogfooding. Whatever the company needed, I figured out how to build it. Data Debug went from an idea to 30+ regulars & 200+ members in the community. I coordinated a 5-company Data Valentine challenge across Recce, Greybeam, dltHub, Database Tycoon, & Bauplan. Co-hosted the Data Renegades Podcast & sat across from folks like Simon Willison, Wes McKinney, Roger Magoulas, Benn Stancil, & Maxime Beauchemin. Built a partnership program with consulting firms that drove real pilots. None of this was solo. CL gave me room to build and trusted my instincts on how we should talk about what Recce does. Karen Hsieh was a constant thought partner on events, content, & getting things out the door. Jenneviere Villegas - community, events, GTM strategist's mentorship & guidance on the content side shaped how I think about building content systems. The whole team in Taiwan welcomed me with incredible food & real generosity every time I visited. Data Debug SF isn't going anywhere. I'm not going anywhere. I'll have more to say soon about my next chapter, what I built here, & what I learned about using AI to engineer a go-to-market function. More of that on my personal blog too. For now, if you're in the data community in SF, come through to the next Data Debug. We still have some April speaking slots open :)
-
史家瑞 已按讚史家瑞 已按讚🤗 Big thank you to Hamzah Chaudhary, co-founder of Lightdash, for flying all the way from London to Taipei for two back-to-back events, and for the incredibly thoughtful custom Taiwan 🧋 🇹🇼 stickers. 💜 And to Shanzé Ijaz Munir, Vincent Chen, LI KUAN LIAO, Laurence Chen, Allen Wang and dbt Labs for making it all happen. 📍 數創實驗室 DataInnoLab: Hamzah walked us through the evolution of dashboards — from the early text-to-SQL hype, to Gen BI, to what he calls "Agentic BI." A fundamentally different approach where agents handle the repetitive work, and your data team finally gets to do what they're actually good at. 📍 Taipei dbt Meetup: Live demo. No recording, no polish theater. Just real debugging, real questions, real vision made visible. The kind of talk where you leave thinking "this is actually happening." My biggest takeaway: The data team's job isn't to answer questions anymore. It's to maintain the semantic layer and let agents do the rest. Data engineers become architects. 80% of ad-hoc requests get handled without touching your best people. His example explains it: let your Michelin chef cook for VIP customers and give everyone else a great recipe so they can make their own sandwich. 7 days to migrate a whole team on agentic BI. One command to move from open source to cloud. That's not a roadmap, that's already real. Nobody hand-writes SQL anymore. It's time to bring agents into your data work. #dbt #DataEngineering #AgenticBI #LightDash #dbtTaipei #DataCommunity
-
史家瑞 已按讚史家瑞 已按讚You ship a dbt change. Three days later, your CFO asks why the forecasts are off. Or worse: some automation consuming your data quietly breaks. An agent makes decisions on bad numbers. Data isn't just for your team anymore. It's feeding agents and automations. Now coding agents are making data pipeline and model changes. Reviewing and validating becomes the bottleneck. You can't move fast with confidence if you don't know what changed and what it means. I'm excited to share what we've been building at Recce. We expected one change in this demo PR. Recce found two and surfaced them for human review. Without a data review agent, you either discover issues too late, or install rigorous workflows that slow everybody down. With Recce, you catch it in the PR. Your team moves fast with confidence. Have you dealt with this? Comment below or try Recce at reccehq.com. #DataEngineering #dbt #DataQuality
-
史家瑞 已按讚史家瑞 已按讚I've been playing around with letting LLMs own setting up models. A lot of planning, custom skills & relevant MCPs, then seeing what happens. So watching Aldrin M. from bauplan do this with an entire data pipeline for our final Data Valentine Challenege Week was cool. He set up a Claude file, a readme & a narration directory with task lists for each step, then told the agent "start the story." The agent built a 3-step pipeline from scratch: ingest raw satellite telemetry, run validations, then a full WAP workflow. No code written by hand. The fun part was watching the agent hit real walls. It hallucinated imports, tried to run directly against main (bauplan blocks that unless you use dry-run), kept defaulting to pandas instead of pyarrow. Aldrin's note: "it seems like it's better to explicitly say don't use pandas rather than just encouraging other libraries." If you've prompted an LLM to write code, you know. bauplan's transactional branches are what made this safe to run. Iceberg on S3, git-for-data catalog. The agent does all its work on a branch, validations run as expectations inside the pipeline. If everything passes, merge to production. If not, the branch is disposable. Aldrin's closer: "I didn't have to do any programming myself." Check out the full replay of this & all the Data Valentine Week sessions on Recce's youtube! #DataValentines #DataEngineering
-
史家瑞 已按讚史家瑞 已按讚Chloé Anderson pulled up Stephen Sciortino's dbt lineage graph and immediately: "She's trying her best, but honestly? This is giving overwhelmed and overworked." Data Valentines day 4 was Database Tycoon LLC doing a full dbt makeover, reality TV style. Stephen brought his NYC bus analytics project & Chloe started auditing everything that couldn't justify itself. Duplicate staging models pulling from the same source. An intermediate model doing a cross join, full Cartesian product of every bus stop with every route, that nothing even used. Stephen: "I thought I might need it for something." Chloe: "You're not gonna need it." Stephen: "Less is more, heard." Then the unused dimensions. Dim Date, Dim Burrow, Dim Day Type, all with zero downstream dependencies. 7 models hanging off one source, 3 leading nowhere. Even MetricFlow was in there, added "in case I need it later." Chloe's mom's rule for getting dressed: always take one accessory off. Then take another one off. Same energy. They deleted it all, regenerated dbt docs, and the lineage went from chaos to clean. It can be hard to delete what's unnecessary though. Sometimes it's because you don't know what it will break downstream, or why it was added even if nothing depends on it. Maybe you'd need certain tables. It's easy to assume someone built them with something in mind, even if that someone was you 6 months ago and you can't remember. Chloe's closer: "The best code is the code you don't write. Or in this case, the code you delete." Delete fearlessly. Git remembers everything. Last day tomorrow: Bauplan's "Let AI Build Your Pipelines Without Breaking Your Heart (or Production)" Links in comments
-
史家瑞 已發表感受史家瑞 已發表感受Day 3 of Data Valentine Challenge is done!! dltHub walked through their workspace workflow. Scaffolds for pulling from REST APIs (they used GitHub in the demo), one init command to spin up a project, and cursor rules + source-specific YAML so an LLM can fill in your config without you having to live in the API docs. Then Marimo notebooks and Ibis to explore the data: write SQL or Python in the same notebook, same abstraction over the backend. Commits per month, commits per contributor, a couple of Altair charts. Pipeline from zero to running without writing the pipeline code yourself. Having templates and LLM-friendly context so you're not hand-holding pagination and rate limits is a real time-saver. dltHub's been in the ecosystem a while and it shows in how they've structured the scaffolds. Join us tomorrow with Database Tycoon LLC's From Hot Mess to Happily Ever After: A dbt Glow-Up Links in comments.
-
史家瑞 已按讚史家瑞 已按讚💘 Fall in love with your data stack this Valentine's week. We're hosting the Data Valentine Challenge: 5 days, 5 companies, 5 sessions that fix real data problems. Feb 9–13 | 9am PT daily | Virtual & free The lineup: Mon | Relationship Advice for You and Your AI Agents (Recce) Prompting techniques that make your agents actually listen — measured against ade-bench. Tue | Dear Snowflake, I Want an Open Relationship (Greybeam) Query Snowflake + Google Sheets + parquet in one DuckDB workflow. No pipelines needed. Wed | Pipelines That Don't Ghost You (dltHub) Build ingestion pipelines that show up when they're supposed to. Thu | From Hot Mess to Happily Ever After (Database Tycoon LLC) Give your broken dbt workflow the glow up it deserves. Fri | Let AI Build Your Pipelines Without Breaking Your Heart (bauplan) Scaffold safe pipelines with Claude Code + git-like branch isolation. Can't make it live? Every session is recorded. Attend all 5 and we'll mail you a data valentine sticker 💗 Register → https://lnkd.in/g8CxpAbz #DataEngineering #DataQuality #DataOps
-
史家瑞 已按讚史家瑞 已按讚Data engineers are so buried in pipeline maintenance they cannot generate insight. The complexity is consuming the value. Hear Roger Magoulas break this down on the latest Data Renegades Podcast episode with CL Kao & Dori Wilson — available on YouTube, Spotify, Apple Podcasts, or wherever you get your podcasts. #dataengineering #datateams #analytics #dataops
-
史家瑞 已按讚史家瑞 已按讚How Mode started from an internal tool at Yammer. Benn and his team built a SQL query editor in a browser with charts on top. No market research. No competitive analysis. They just built what they needed. When they looked around, Facebook, LinkedIn, Airbnb, and Uber had all built similar internal tools. That was the validation. Sometimes the best products come from solving your own problem first. #StartupStory #ProductDevelopment #DataTools
工作經歷和教育背景
-
Recce
專案
-
Term Structure Corporate Website
Provide Term Structure a Platform to Communicate to Prospective Customers
Beginning April of 2024, Term Structure released their application to the public. While this was an exciting time, the adoption rate of new users was lower than expected. I worked with the marketing team and we identified the lack of a compelling company website as the main culprit. Over the period of a week in July, I revamped the existing website, building out a comprehensive solution using NextJS and Tailwind…Provide Term Structure a Platform to Communicate to Prospective Customers
Beginning April of 2024, Term Structure released their application to the public. While this was an exciting time, the adoption rate of new users was lower than expected. I worked with the marketing team and we identified the lack of a compelling company website as the main culprit. Over the period of a week in July, I revamped the existing website, building out a comprehensive solution using NextJS and Tailwind. One of the additional benefits of the rebuild was aligning the design and implementation to be the same as our application and internal tools. Not only did this create a friendly experience for potential users, but allowed our development team to more easily support the website in the future. -
Minesweeper, reimagined
In an attempt to build on my understanding of React and the React ecosystem, I undertook a 5-day challenge to recreate a classic Windows game, Minesweeper. I may have gotten a bit loopy while developing it, as I took some liberties (I called it, "Mimesweeper"... you'll understand why when you play it 😀 )
This project makes use of the most recent version of React , using TypeScript as well as the Functional Programming paradigm. I made use of retro font's to give it a throwback aesthetic. -
Day Calendar
- 目前
查看專案A simple layout engine that works with any number of overlapping events. Designed to be robust and efficient.
-
Write Wall for Google Chrome
- 目前
查看專案Write Wall is an attempt to create a very simple to use way to share a clipboard between multiple computers running under the same Google profile. Think home/work or even two work computers when you want to copy/paste simple text between the two.
語言能力
-
Japanese
中高級（商務會話）
-
English
母語
-
Chinese (Mandarin)
中級（日常會話）
推薦已收到
3 人推薦史家瑞
加入即可瀏覽