We scheduled it from 6 to 9 PM. By 10:30, we were prodding people (gently) to clear the building so the venue could close. We still had a decent sized group though wanting to continue the conversations, so we walked over to a local restaurant.

This is the lesson community keeps teaching me. Content gets people through the door. But it’s the connections they make that convinces them to stay.

That’s exactly why co-hosting with CVC Taiwan worked so well. CVC brought in entrepreneurs, alumni, industry veterans, and investors who don’t usually attend AI-only events. Mixing them with our regular crowd of AI builders, founders, operators, and researchers created the kind of cross-pollination you can’t force. It made for more valuable connections and engaging discussions.

The talks delivered.

Chan-Liang Wu from MICROIP covered what it takes to run modern AI workloads on embedded hardware: the tradeoffs across performance, power, cost, and silicon area that make or break edge AI products. Silicon design discipline applied to AI is a perspective most software-only AI conversations skip entirely.

Tek Lim from Next Integral not only gave a talk, he brought a live demo. He set up a computer vision rig on a table with sensors checking whether a bottle cap was fully closed, simulating quality control on a bottling line during liquid filling. Seeing it run on a table beats reading about it on a slide every time.

Then came the pitches. Seven founders shared what they’re building, across a range a diverse range of use cases from music, HR, safety, business cards, and AI agents:

• Jasmine Ow at Sonicstride

• Oumnya Benhassou at CloudStation

• Timur Aiusheev at Luma (changed from 1 Percent)

• Elisa Chang at Digital Business Card

• Tony Chi at Xybervu Veloscope

• Jih-Ming Chen at SkipInterview

• Christopher Pribadi at Bysik AI

After the official program ended, the crowd didn’t drift toward the exit. People exchanged business cards and LinkedIn QR codes. Founders cornered the team to ask about volunteering, content collaborations, presenting at future events, and bringing colleagues to future events. That’s what led to our moving to a new after-event venue!

That post-event drift is a clear sign for us that the community is coming together. We are becoming a genuine community that people value and schedule around, not just another AI event.

If you were there, what made it stick for you? Please let us know, we would love to hear your thoughts on what you liked and what we can improve.

We’re still processing video from the talks and will share more from each speaker in a the next newsletter. For now, huge kudos to the team that pulled this off: Mark Lin, Hamza Tahiri, Luna Luovula, Jordi Vallejo, and the day-of event crew of Hollis Cutler, Diego VG, Leo 何霆鋒, Ben Sadeghi, and Gareth Kennedy. They did the heavy lifting and it showed.

Our next event is coming later in May (more announcements to come). If you want to get involved with logistics or content planning, reach out and we can share where we need help most.

See you soon,

Mark & the TASC Team

All links are now fixed below. With our Startup Networking Night: Embedded AI In Practice event happening this Wednesday (tomorrow), here’s a closer look at our two speakers and co-host, so you know who you will be listening to (and who to corner with questions afterward).

Chan-Liang Wu

CTO & GM of IC Design BU, MICROIP

Chan-Liang’s career covers a wide range of hard problems in modern silicon, including high-voltage driver ICs, FinFET IP development, SoC design, and 3DIC integration, with earlier leadership roles at M31 Technology and Himax before joining MICROIP. He has also run the operational backbone of an IC design house (CAD, EDA, ESD, device teams, and license and tool procurement), which gives him a pragmatic read on what production-ready really costs. At MICROIP’s 2025 Grand Technology Seminar, he introduced AIVO, an edge AI platform that applies chip-design principles such as modular IP, visual composition, and reusability to make AI development more structured. AIVO went on to anchor MICROIP’s CES 2026 presence, with deployments across transportation security, agriculture, and autonomous systems. His talk offers a practical perspective on the real performance, power, and cost tradeoffs behind running modern AI workloads on embedded hardware.

Tek Lim

Founder & CEO, Next Integral

Tek is focused on a costly, under-addressed problem in manufacturing. Most factories still rely on manual checks and a patchwork of disconnected tools, and the resulting defects, rework, and recurring losses add up year after year. His company Next Integral builds NEXUO, an edge AI vision platform that combines computer vision and real-time analytics on-device. It detects defects and process deviations early, standardizes inspection across lines and sites, and is positioned as a more affordable alternative to traditional AOI systems for industrial SMEs, with payback often inside seven months. Tek holds a PhD in Physics from École Polytechnique and has 15+ years of experience bridging AI research and real-world deployment. NEXUO began in Food & Beverage packaging, covering filling level, cap integrity, label alignment, and date/lot/barcode readability, and is expanding into broader packaging QA, logistics, and high-throughput visual workflows. The company is backed by Taiwan Tech Arena (TTA) and has been featured at Smart City Summit & Expo Taipei, Meet Taipei, SWITCH Singapore, and CES 2026. His session covers CV fundamentals, design tradeoffs, real deployment challenges, and what tends to scale once these systems move from prototype into production.

Columbia Venture Community Taiwan

Global Columbia alumni network for entrepreneurship & innovation

Founded in 2006, Columbia Venture Community (CVC) is a 7,500+ strong global network of Columbia University alumni, students, and staff working across every corner of entrepreneurship and innovation. CVC runs regular programming in San Francisco and New York, with active chapters and events spanning 15+ cities from Silicon Valley to Seoul, including a Taiwan chapter that plugs local founders, operators, and investors into that worldwide pipeline. It’s a useful bridge for anyone building between Taipei and the U.S. ecosystem, whether you’re raising, hiring, partnering, or just looking for a smarter room. The global community has hosted 500+ events and counts unicorn founders and prominent VCs among its alumni speakers.

See you Wednesday and please remember to register ahead of time!

Mark & the TASC Team

Did you know we have a Facebook page for the Taiwan AI Startups Community? We just recently launched it as a place for both Chinese and English content and event announcements. Come join us there!

First, we’re co-hosting with Columbia Venture Community Taiwan - a monthly gathering of entrepreneurs, alumni, industry veterans, and investors active in Taiwan’s startup scene. Joining forces means a lot more cross-pollination between AI builders and the startup and investor community, and a good excuse to stay late!

Second, we’re focusing on the part of the AI stack most events skip: embedded AI.

Most AI conversations stay at the model layer. What runs on a GPU in a data center. What comes out of an API. What writes the best code. But there’s a whole other world where AI actually lives — on a camera at a factory line, on a sensor inside industrial equipment, on chips small enough to fit in devices running off a tiny power budget. Taiwan has deep expertise in this layer, and we wanted to dig in.

We brought in two speakers coming at it from opposite ends:

• Chan-Liang Wu — CTO & GM of IC Design BU at MICROIP. His career spans CAD/EDA, SoC development, 3DIC integration, and IP development for FinFET processes. Chan-Liang brings the silicon-first view: what it takes to run modern AI workloads on embedded hardware, and how performance, power, and cost actually trade off at the edge.

• Tek Lim — Founder & CEO of Next Integral, building edge-based computer vision for industrial applications. PhD in Physics from École Polytechnique, 15+ years bridging AI research with real-world deployment. Tek’s talk runs from CV fundamentals through what it actually takes to ship these systems into production: the design tradeoffs, the deployment headaches, the lessons from running at scale.

One focuses on applied computer vision. The other dives deeper into the hardware that makes it possible. Together you will get a real picture of what shipping AI from research to live production systems looks like.

After the talks, we’ll move into startup pitches, founder breakout sessions, and open networking.

• When: Wednesday, April 29, 6:00-9:00 PM

• Where: FutureWard Central, Songshan District, Taipei

• Cost: NT$700, cash at the door

• Who should come: Founders, engineers, researchers, and investors working on applied ML, computer vision, or anything embedded

• Sign up here: https://trib.co/embeddedaitalk

If you know someone in Taiwan building or investing in this space, please pass this along!

See you this Wednesday,

Mark & the TASC Team

The first scenario hit on something many companies are quietly doing: repurposing customer data collected for one purpose (say, marketing) to train AI models. Matthew made it clear that under the GDPR’s purpose limitation principle, this is a serious problem — and not just in Europe. Privacy laws modeled on the GDPR are spreading globally, from Taiwan’s recently amended Personal Data Protection Act to other frameworks across Asia and beyond. The penalties aren’t just fines. He pointed to the concept of “algorithmic disgorgement,” where regulators can force a company to delete not just the data, but the entire AI model trained on it!

From there, he moved into the murky territory of AI hallucinations — what happens when a chatbot confidently gives a customer wrong information that leads to real financial losses. Matthew broke down the dual legal exposure: compliance risk (regulatory fines) and liability risk (lawsuits from harmed parties). He then walked through a third scenario involving adversarial cyberattacks on AI systems handling sensitive health data, tying together privacy, information security, and the special protections required under laws like the GDPR for health-related personal data.

So what’s the solution? Matthew made a compelling case for building your governance structure around three international standards: ISO 42001 for AI management systems, ISO 27001 for information security, and ISO 27701 for privacy. Together, these create what he called a “defensible audit trail” — evidence that your organization exercised reasonable care. This matters more than ever, he argued, because insurance companies are increasingly excluding AI-related losses from coverage. If you can’t insure against the risk, the next best thing is proving you did everything reasonable to prevent it.

He closed with a broader look at the regulatory landscape, including the EU AI Act’s extraterritorial reach and the political forces shaping AI law globally. His takeaway was clear: the regulatory environment is evolving fast, and organizations that get ahead of it with a standards-based governance framework will be in a far stronger position than those scrambling to react. For anyone navigating the intersection of AI, law, and risk, this was one of the most practical talks of the evening.

If you joined us for Matthew’s talk, what were your big takeaways from the session?

The team is currently organizing our next event for later this month. If you are interested in getting involved in either event logistics or content planning, please reach out to us and we can share more about the areas we need most help with.

Thanks again and see you later this month!

Taiwan AI Startups Community team

If your AI product caused a privacy breach tomorrow, do you know what your liability looks like? Most founders I talk to don’t. That’s not a criticism. The rules around AI are messy, they vary by country, and they keep changing.

But the consequences aren’t theoretical. Companies have already been hit with fines, lost customers, and taken reputational damage from AI failures. And we’re still early.

A few things worth knowing:

• Governments are moving. The EU AI Act is in motion. Other countries are drafting their own versions. And even where specific AI laws don’t exist yet, privacy and consumer protection laws already cover a lot of what AI products do. Not knowing about a regulation doesn’t protect you from it.

• The risk isn’t just legal. One bad AI output that goes viral, one data leak, one compliance gap. Any of these can cost you customers and funding. When you’re early-stage, that kind of hit is hard to recover from.

• Governance frameworks aren’t just paperwork. Standards like ISO 42001, ISO 27001, and ISO 27701 give you a structured way to identify risks and deal with them before they become emergencies. They also make due diligence conversations with investors and enterprise customers a lot smoother.

You don’t have to figure all of this out by yourself.

At our next event, Matthew Seet (who’s written four books on this topic and used to teach international law at NUS) is going to break it down in plain language. No legalese. After his talk, we’ll do an open discussion where founders can bring real questions about the legal and governance side of building AI.

• Time: Thursday, March 19, 6:00-8:00 PM

• FutureWard Central, Songshan District, Taipei

• Sign up here: https://trib.co/taiwan-ai-legal

Spots need approval, so don’t wait. See you there,

Mark & the TASC Team

If you’re building AI products, you’ve probably had that thought at 2am. What happens if something goes wrong? What laws even apply to what I’m building? Most founders skip these questions until a customer, an investor, or a regulator forces the conversation.

We brought in someone who can make this make sense.

Matthew Seet is a Singaporean based in Taiwan (Gold Card holder) who’s written four books on AI risk management for publishers like Springer Nature and Globe Law and Business. He also runs a legal AI risk management edtech company and used to teach international law at the National University of Singapore.

His talk, “Navigating AI Risks and the Evolving Legal Landscape,” will cover the financial and legal risks of building and selling AI tools, how the regulatory environment is changing (explained so non-lawyers can actually follow), and how governance standards like the ISO trifecta help manage risk before it blows up.

After the presentation, we’ll open a conversation focused on founders to discuss questions about shipping AI products, compliance headaches, and the governance stuff nobody warns you about.

• When: Thursday, March 19, 6:00-8:00 PM

• Where: FutureWard Central, Songshan District, Taipei

• Who should come: Anyone building, deploying, or working hands-on with AI

• Registration: Approval required

• Sign up here: https://trib.co/taiwan-ai-legal

If you know someone in Taiwan building AI who’d get value from this, pass it along!

See you March 19,

Mark & the TASC Team

50 people showed up to hear Ben Sadeghi and Gareth Kennedy share how they use AI in their work. What made it interesting was hearing real examples from their daily workflows, not theoretical possibilities.

Some highlights:

Gareth walked through his “automated junior analyst” approach - running the same AI query 10 times, getting a distribution of responses, throwing out extremes, and using the median. Simple statistical method to filter hallucinations.

Ben discussed cross-validation between multiple LLMs because he keeps getting reports with inaccurate information. He uses “LLM reflection” to ask the model to review its own answer as a cheaper verification method. Never trust a single output.

The discussion also dove into how much can we trust AI output. AI has improved, but still prone to errors. We discussed scientific papers becoming “AI slop” when peer reviewers use AI to review AI-written papers. Humans need to verify, and not blindly accept whatever AI generates.

What’s next:

We’re taking February off for the upcoming holidays and planning for our next event in March (and for the next few monthly events beyond). We need help in three areas:

Venue - Need space for 100 people in Taipei. We turned away half of the 120 people who registered. We clearly underestimated demand!

Speakers - Looking for people building or using AI to share practical workflows and insights, especially Taiwanese speakers and women in AI/tech. Raise your hand if that fits you or send over some recommendations for awesome speakers.

Volunteers - Community doesn’t scale without people who can help. We need support with event logistics, speaker outreach, promotion, and other behind-the-scenes work. Reply to this newsletter if you’re interested.

Huge thanks to Evy Chang, Mark Lin, and Luna Luovula who made Thursday happen - venue sourcing, speaker coordination, registration management, event greeters, and everything in between.

If you were there, thanks for showing up and bringing good questions. If you missed it, we’ll have more space in March.

See you soon,
The TASC Team

We are doing something a bit different from what we have seen with other AI events. Taiwan has plenty of product demos, high-level networking events, and deep technical talks. What we don’t have are events where speakers open up about how they use AI and share "here's where it works, here's where it doesn't."

So we invited two speakers to bring different perspectives and experiences:

• Ben Sadeghi - Ex-YC founder/CTO building omoomi, an offline-first, privacy-aware AI companion. He’ll share how he uses AI as a thinking partner when building products, not just for automation.

• Gareth Kennedy, PhD - Scientist who worked across Australia, Spain, Germany, and China before moving into industry AI. Led MLOps and AI governance teams at Deloitte Australia while keeping roles at universities and an astrophysics research center. Now building a startup around AI in scientific research.

We will kick off with the panel discussion followed by networking.

Thursday, January 29, 6:00-8:00 PM
Xinyi District, Taipei

Sign up here: https://trib.co/ai-experts

One ask: If you know founders, engineers, or researchers in Taiwan who’d get value from this, please share with them. We’re building this community for people actively using AI in their work, and your network has exactly those people in it 😁

See you next Thursday,
Mark & the TASC team