Home of Singapore's AI Ecosystem Grow with us: https://xitter.cf/t.co/0oeTD9rHrD

One-North, Singapore
Joined January 2025
Using AI tools at work but never had a cyber health check? Eligible Singapore #SMEs can get up to 70% co-funding for CSA’s CISO-as-a-Service: an outsourced security lead who checks your gaps and preps you for Cyber Essentials. Details below 👇 #Cybersecurity #AIinFinance
1
51
A medium-sized app can surface 6,000+ vulnerabilities. A @MAS_sg team built RiskVibe to triage them: 1,600 findings down to 200 for closer review, and about 3,000 man-days saved a year. The final call still sits with people.
1
52
How @MAS_sg ’s RiskVibe handles vulnerability scans, in 4 steps. AI drafts. People decide.
1
20
Models change. How do you know your AI tool is still giving good answers? The @MAS_sg RiskVibe team checks outputs against 30 expert-curated “gold samples”, with an LLM as the judge, every time something changes. A simple habit worth borrowing.
14
Find Lorong AI and our members this week: @ConDaysEU : booth A22, Suntec, 27–28 Oct @SwitchSingapore : booth T01, Sands Expo, 27–29 Oct Come meet the team and some of the people building #AI in #Singapore.
1
113
What does a room full of aviation leaders and AI builders look like? ✈️ At our last MAP: Aviation Edition, CAAS, SATS and @SingaporeAir shared 6 challenge statements, then builders pitched to them on the spot. The energy was sky-high. More to come. Follow us to stay updated.
1
1
63
Buying an AI or digital tool with a grant in Singapore? Since 30 Sep, PSG and EDG applications go through EDGE. One rule to remember, per Enterprise Singapore: apply before you pay. Work started or deposits paid before you submit make it retrospective. #EDGEGrant #LorongAI
2
67
Thinking of buying an AI tool for your business? Answer three questions first: 1. Which task takes too long or goes wrong too often? 2. How will you measure the change? Note today’s number. 3. Who will use it every day? Then talk to a vendor. #SMESingapore #LorongAI
59
Lorong AI retweeted
Another fruitful night at @Lorong_AI! 🦌✨ More open models should mean more possibilities, not more complexity. Yesterday, we explored how builders can seamlessly turn AI experiments into production-ready applications. With @Bitdeer_AI, developers get easy access to open models plus total control over deployment, scalability, security, and human-in-the-loop workflows. 🛠️🚀 Huge thanks to our speakers and everyone who got hands-on with us! 🎉 Next stop ➡️ #Malaysia Roundtable! 🇲🇾 See you there 👉 luma.com/i7xgaape 💡 #Malaysia's AI Advantage : From Infrastructure to Innovation 🗓️ Wednesday, 28 Oct 📍 Pullman Kuala Lumpur City Centre Hotel & Residences
3
2
1
19
2,284
Ten turns to answer one question. Zuno Carbon shared at Lorong AI how their agent kept rewriting SQL. The fix: replace the model-written query with a deterministic function the agent calls. If a step must be exact every time, don’t leave it to the model. #AIAgents #LorongAI
1
2
101
Hiring AI talent in Singapore? Three places to start: AI Singapore: talent matching and a jobs portal EDB: introductions to partners and government support Lorong AI: meet practitioners and researchers in person Links below. #SingaporeAI #LorongAI
1
1
2
127
Lorong AI retweeted
How do you keep AI accountable in a bank? Ivan Poon (Standard Chartered) shared at Lorong AI: combine retrieval, break tasks into smaller steps, add evaluation and include human review so every output is grounded in evidence. Test: Can a reviewer trace each answer to its source
1
1
2
88
How do you keep AI accountable in a bank? Ivan Poon (Standard Chartered) shared at Lorong AI: combine retrieval, break tasks into smaller steps, add evaluation and include human review so every output is grounded in evidence. Test: Can a reviewer trace each answer to its source
1
1
2
88
We asked builders and business owners the same question about AI, and got very different answers. Builders: making AI perform reliably, at up to 52%. Founders and business owners: 67% said fitting AI into how work gets done, or getting people ready for it.
1
56
Builders aren't blind to this. At our session on AI agents, 73% said their biggest open question, once an agent passes its evals, is a simple one: will it hold up with real users?
1
15
So the gap may sit between the people building AI and the people expected to use it. One way to close it, from Timothy Lin of Resaro: test with real scenarios from the team that will use the tool. Changing an agent's setup alone can shift performance by 15+ points.
7
Building #AI for aviation? Tomorrow, hear the problems straight from the people who own them! MOT on the AI mission, then SATS on their challenges, plus demos from Overhaul, Cabinsense and Cohere. Tue 13 Oct · 2.30pm · Lorong AI @ one-north
1
60