Lytx

I am pleased to have served as a judge during the Innovate to Grow event (i2G) at UC Merced. The talent displayed by the next generation of engineering students is truly impressive. University of California, Merced, #ucmerced , #materialsengineering , #bioengineering , Amanda Day, Ashlie Martini, Kara McCloskey, Sarah Kurtz, #ucmercedengineering , #jackoswald , #edwardkolodka , #shawnnewsam

I’m proud to have joined our San Diego team in giving back locally and to see employees around the world supporting causes that matter to them. Through our Volunteer Time Off program, every Lytx, Inc. employee has the opportunity to make an impact in the communities they care about most.

Nobody talks about the part that actually makes enterprise AI safe. Not the model. Not the prompts. The governance layer. You can deploy the best LLM in the world. But if you have not answered these five questions first, you have not deployed AI. You have deployed risk. 1. Who can trigger an AI action and who cannot? 2. What can the AI change without human review and what requires approval? 3. When the AI is wrong, who is accountable ? 4. How do you audit what the AI decided and why? 5. When do you pull the plug and what is the rollback plan? Most enterprise AI discussions stop at question zero: "which model should we use?" But teams that deploy AI safely spend far more time on governance than on the model itself. It is the part that determines whether your AI system is still trusted and still running, six months from now. #AIGovernance #GenerativeAI #EnterprisAI #AgenticAI #ResponsibleAI

Time to rock and roll .. Commercial Vehicle Show 2026!!!

Oil and gas fleets operate in some of the highest‑risk environments on the road. This blog breaks down how safety and tracking help leaders reduce risk, protect drivers, and keep operations moving in unpredictable conditions. #FleetSafety #ClaimsManagement #IndustryInsights

Everyone told me LinkedIn comments were just noise. So I tested it anyway. Tracked 400+ hiring posts in 30 days using a simple signal system. Found 18 hiring managers actively struggling in comment sections. Started conversations with all 18. Closed 6 roles directly from it. Here’s what nobody talks about: LinkedIn comments aren’t engagement. They’re hiring intent. Here’s the system I built: Step 1: Find hiring / “we’re scaling” / “we can’t find talent” posts Step 2: Scan every commenter for signals (hiring, team growth, frustration) Step 3: Qualify — founder, HR, or decision-maker Step 4: Extract their pain from what they actually wrote Step 5: Send one message referencing their exact comment Step 6: Track replies (simple notification system) Step 7: Continue the conversation manually (not automated) That’s it. 7 steps. Zero cold pitching. The hard part wasn’t building the system. The hard part was realizing this: My best clients were never in a database. They were in comment sections. Publicly explaining their hiring problems. For free. If you’re a recruiter: Stop chasing leads. Start reading comments. That’s where intent lives. What’s one manual task in your recruiting workflow you wish you could automate? 💬 Comment "RECRUIT" and I’ll DM you the full setup.

Huge thanks to Gareth Danks for the professional headshots (oh and some not so, lol). Seriously we had so much fun and yes guys this is us, team Reed! I’m so proud that I created a business that has had the opportunity to bring my husband and son (my two best friends) in to help and support me. Honestly I don’t know how I would be able to manage my day to day without them both. #familyrun #supportlocal #dreamteam #inittogether #trust #honesty

Apparently, this post became the final trigger for me to finally formulate my position on this whole “manual QA is doomed” narrative. :) Not because it is unique or even worse than the endless stream of similar posts flooding LinkedIn for years. It just happened to appear at the exact moment when my mood and half a liter of Spaten aligned. ;) Funny enough, one thing AI has clearly improved is people’s ability to produce polished, structured, and confident-sounding generic takes. To be fair, for simple website UI checks, straightforward unit-level validation, and other highly formalized tasks, manual QA is becoming less central. But that logic breaks down fast in embedded systems, hardware/firmware, medical devices, defense, automotive, industrial systems, telecom, and other safety-critical domains, where quality depends on human behavior, physical reality, ambiguity, integration boundaries, timing, and failure modes no spec fully captures. AI can reduce manual execution. It cannot replace human judgment, imagination, intuition shaped by lived experience, or the ability to evaluate products as a human being using something built for other human beings. And! Special thanks to ChatGPT for helping me express that clearly, structurally, and without spelling mistakes. ;)

There are many concepts in software engineering you don’t truly understand until you’ve implemented or debugged them yourself. Take concurrency and parallelism for example, you can read about them endlessly, but real understanding only comes when things break. Instead of waiting to encounter these problems at work, try this: a. Pick an open-source project you find interesting b. Search for keywords related to the concept you want to learn c. Find relevant issues d. Work on them and submit a PR It’s a far more intentional and faster way to grow than waiting to naturally encounter these concepts at your day job.