iLabService

Autonomous labs are coming. But I'd like to ask: who watches the freezer at 2:14 AM? The conversation around autonomous science is getting bigger: AI-designed experiments, robotic execution, higher throughput, 24/7 operation. Good. But autonomy does not end when the robot completes the protocol. At 2:14 AM, the lab may still need to know: 🧊 Is the freezer recovering or failing? 🧪 Which samples are exposed? 🚪 Was the door opened? ⚡ Did the power pattern change? 📣 Who was alerted? ✅ What evidence will QA see on Monday? A lab can automate experiment execution and still depend on manual response when the physical environment changes. That is the uncomfortable gap between an autonomous experiment and an autonomous lab operation. The lab runs for 168 hours. The team is not there for all 168. iLabService focuses on the operating layer around the science: physical signals, asset and sample context, response ownership, and audit-ready evidence. Because a robot running the experiment is only part of the story. The other part is knowing what happened when nobody was standing next to it. We explored this in the 168-Hour Lab: https://lnkd.in/gXq4zhJx #LabOps #AutonomousLabs #LabAutomation #DataIntegrity #EHS

Thrilled to share that I have recently joined iLabService as their Business Development Consultant – focused exclusively on supporting their market expansion into the US. 🤝 I truly appreciate Kang Li and Tony Chung for their trust and the opportunity to collaborate. I am also grateful to everyone who has supported me along my journey in the life sciences industry – your insights, encouragement, and partnerships have been invaluable! Here is the new challenge I am excited to help address: Laboratory operations are continuous. Critical events can occur at any moment – and when they do, the priority is clear: rapid alerting, accurate documentation, and auditable evidence. With over 200+ clients since 2016 and backing from Y Combinator, iLabService's AIoT-powered LabOps intelligence platform delivers real-time detection, documentation, and alerting – helping labs improve efficiency, reduce waste, and achieve audit-readiness. It has been a pleasure working closely with the iLabService team – their attention to detail, thoughtfulness, and collaborative approach have been evident from day one. The platform is intuitive, well-designed, and highly optimizable to solve a range of operational bottlenecks that I am here to help US labs navigate – whether it is understanding the platform, exploring a pilot, or discussing opportunities to simplify your LabOps. ✅ Follow iLabService for insights on compliant, audit-ready LabOps – and if you have a lab operational challenge in mind, I would love to hear about it. Feel free to DM me, especially if we haven't caught up in a while! 💬 #LabOps #RegulatoryAffairs #FDA #GxP #LifeSciences

😇 Grateful for a meaningful start at iLabService! To kick off this new chapter, I am fortunate to have contributed to a successful solution deployment: Lab Inventory Control Solution for the QC laboratory of a renowned drug manufacturer at HKSTP Tai Po InnoPark. By automating inventory management, this solution directly empowers the QC team by: ⏱️ Saving valuable time and reducing manual workload of the microbiologists and chemists. 📈 Ensuring accurate stock information with automated data tracking. ⚠️ Minimizing operational risks through real-time alerts. 🔮 Providing future upgradability to support broader digital transformation, experiment & safety monitoring, and instrument management. A Huge Thank You to the Team This deployment was a true collaborative effort. Deeply grateful for the dedication of the Hong Kong Labware team - Johnny Chiu, Michelle Lee, Roy Lee, Howard Mak, and my colleagues Zephyrus Zhang and 石生伟. Global Expansion: Launching into North America 🌎 I am grateful to have confirmed partnerships with two professional consultants to jointly explore the North American market.  It is made possible by their trust and the generous support of my supervisor, Kang Li. Stepping into this new territory from the ground up marks an exciting new chapter for both myself and the company. Special thanks also to my friends at the HKSTP Partnerships and HKSTP Global Connect teams. Your incredible ecosystem support means a lot to us and is vital in helping our pipeline grow healthier every day. I feel encouraged and inspired that more success stories will happen in the near future, as we continue to empower laboratories and create a meaningful impact together! 🔬✨ #LaboratoryAutomation #QC #PharmaManufacturing #HKSTP #BusinessGrowth #DigitalTransformation

Your lab is staffed for 50 hours a week. Your samples are exposed for 168. That gap is bigger than most LabOps plans admit. A freezer does not wait for Monday morning. A cryo tank does not care that the team has gone home. A gas cylinder does not schedule its low-pressure moment during business hours. A door opening, power issue, oxygen risk, temperature drift, or compressor warning can happen at 2:14 AM on a Saturday. And when it does, the question is not only: “Did we have monitoring?” The real questions are: 🌙 Who was alerted after hours? ⏱️ How long did the exposure last? 🧊 Which samples or assets were affected? 👤 Who acknowledged and responded? 🛠️ What action was taken? ✅ What evidence can QA or EHS review later? This is the 168-hour problem. Most labs are managed around the staffed week. Risk operates on the full week. That is why after-hours LabOps cannot depend on manual checks, screenshots, and Monday morning reconstruction. If a critical event happens outside normal coverage, the response context and evidence need to be created while the event is happening. Because the other 118 hours still count. #LabOps #LabSafety #DataIntegrity #EHS #DigitalLab

If your lab needs three people, six screenshots, and a Friday afternoon memory hunt to explain one alarm... That is not audit readiness. That is evidence reconstruction. And it happens more often than most teams want to admit. After an event, the search begins: 📸 screenshot from the device portal 📊 export from the monitoring system 🧾 paper log from the room 💬 Teams message from the responder 📧 email from the service vendor 🧠 memory from the person who was there Individually, each piece may be useful. Together, they are fragile. Because QA does not only need to know that something happened. QA needs to know: ⏱️ when it happened 📍 which asset or sample was affected 🚨 who was alerted 👤 who acknowledged it 🛠️ what action was taken ✅ who reviewed and closed it That timeline should not have to be rebuilt after the fact. It should be produced by the operation while the event is happening. This is the difference between monitoring and audit-ready LabOps. Monitoring says: something happened. Operating evidence says: here is the full context, response, ownership, corrective action, and review trail. The future of regulated lab operations is not more screenshots. It is evidence that is born with the event. #LabOps #DataIntegrity #QualityManagement #GxP #DigitalLab

The AI agent says: “Run the experiment.” The lab says: 🧪 The sample is still waiting. ⚠️ The reagent is in inventory, but not in the right condition. 🟡 The instrument is idle, but not actually ready. ⏱️ The queue changed 20 minutes ago. 👤 The operator made a small workaround. 📝 The handoff happened in a hallway. None of that is in the protocol. None of that is in the final data file. But all of it can change what should happen next. This is the part of AI-native labs we do not talk about enough. AI can read papers. AI can write protocols. AI can suggest experiments. AI can analyze results. But if it cannot see the physical state of the lab, it is reasoning over a clean version of a messy system. And labs are not clean digital objects. They are moving samples, aging reagents, busy instruments, interrupted people, local workarounds, environmental drift, and exceptions that appear between steps. So the question is not only: “How smart is the AI?” It is also: “What can the AI actually see?” Because AI cannot run a lab it cannot see. #AIforScience #LabAutomation #DigitalLab #LabOps #AIoT

Bad lab data often starts before the data file exists. By the time a result appears in a LIMS, ELN, CDS, or report, some quality problems may already be baked in. - The sample waited longer than expected. - The reagent was technically in inventory, but not in the right state. - The instrument was available on the calendar, but not actually ready. - The room condition drifted for twenty minutes. - The method was followed, except for one small workaround everyone knows about. - The handoff happened, but the context did not. None of these may look like a “data issue” at first. They look like normal lab life. A queue / A delay / A quick adjustment / A local habit / A missing note / A small exception. But later, when the data is reviewed, investigated, or used to train an AI system, those physical moments matter. Because data quality is not created only at the moment of analysis. It begins earlier with the sample, reagent, instrument, environment, operator, and handoff. This is why the future of lab data integrity is not only better data capture after the run. It is better visibility before the file exists. #DataIntegrity #QualityControl #LabOps #DigitalLab #LabInformatics

Every lab has a shadow operating system. It is not installed on a server. It lives in sticky notes, private spreadsheets, Teams messages, hallway reminders, whiteboards, and the memory of the one person everyone asks before touching that instrument. It sounds like this: 🗣️ “Ask Sarah before you book that slot.” 🗣️ “Don’t use that reagent after Wednesday.” 🗣️“That freezer alarm is usually nothing.” 🗣️“The method says 30 minutes, but we normally wait 45.” 🗣️“The instrument is online, but not really ready.” None of this is necessarily bad. In fact, it is often how experienced labs keep moving. The problem is that shadow systems do not scale. ❌ They do not onboard new people well. ❌They do not travel across shifts. ❌They do not show up in dashboards. ❌They do not create evidence when something goes wrong. And they disappear when the person who knows leaves the room. A lot of lab digitization projects focus on replacing paper. But the bigger opportunity may be replacing invisible coordination. Because many labs are not failing from a lack of effort. They are running on operational knowledge that never made it into the system. #LabOps #DigitalLab #LabAutomation #DataIntegrity

Most labs know what they store. Fewer know what is active. That sounds like a small difference. It is not. A chemical sitting in a cabinet is one kind of safety question. A chemical being weighed, transferred, heated, concentrated, dried, mixed, scaled up, or left mid-process is a very different one. Inventory answers: “Do we have it?” The bench asks: “What is happening to it right now?” That is where a lot of chemical risk quietly changes. Not in the database. Not in the SDS folder. Not in the annual training record. But in the ordinary moment when someone says: “We used this before.” “Let’s make a little more.” “It should be fine.” “I think the limit was around…” This is why chemical safety cannot stop at storage visibility. Knowing what is in the cabinet matters. But knowing what is active at the bench matters just as much. Because risk is not only a property of a chemical. It is a property of a chemical in motion. #LabSafety #ChemicalSafety #EHS #LabOps

A chemical safety failure does not always start with a missing SDS. Sometimes it starts with a material on the bench whose physical risk has not been fully understood. The CSB investigation into the 2010 Texas Tech University chemistry lab explosion is a hard example. A graduate student was severely injured while handling a high-energy metal compound that suddenly detonated in a chemistry laboratory. The CSB later emphasized a point that is still relevant for many research labs: Laboratory safety programs cannot focus only on chemical health hazards. They also need to control the physical hazards of chemicals. That distinction matters. A chemical can have a name. It can have a container. It can have a procedure. It can even have documents around it. But the real question at the bench is different: What is the material actually doing in this form, in this quantity, in this operation, with this person, in this moment? That is where safety becomes physical. Not in the folder. Not in the spreadsheet. Not in the policy. At the bottle. At the bench. At the moment of handling. This is why I think chemical safety has to move beyond document availability. SDS access is necessary. Inventory is necessary. Labels are necessary. But they are not enough if they are disconnected from the actual work. The lab needs to know: What chemical is present? Where is it? How much is being handled? What state is it in? Who is using it? What procedure is being followed? What physical hazard is created by this step? What changed from the last time this work was done? If those answers are reconstructed only after something goes wrong, the system is too late. This is the dangerous gap in many labs. Data exists, but it is not attached tightly enough to the physical reality of the work. The SDS may describe the chemical. The inventory may list the bottle. The procedure may describe the operation. But the risk lives where those layers meet. That is why bottle-level and workflow-level visibility matter. Not as extra bureaucracy. As a way to keep safety context close to the actual handling of chemicals. For EHS, this is about control. For QA, it is about evidence. For scientists, it is about not being left to make critical safety judgments from fragmented information. And for lab leaders, it is a reminder: Chemical safety is not just a document system. It is a physical operating system. The audit may start at the bottle. But the risk starts even earlier, when the lab cannot clearly connect the chemical, the operation, the person, the hazard, and the evidence. #LabSafety #ChemicalSafety #EHS #LabOps #DataIntegrity