Mark Wells

I’ve spent years building technology that solves problems for thousands of people — cloud architectures, AI solutions, serverless systems… all running flawlessly, 24×7. But one day, while deep in a conference call about latency optimization, I got a message from home: "Papa, when will you come play with me?" My 3-year-old daughter didn’t care about cold starts in serverless computing. She just wanted me to start her toy train. In that moment, I realized something powerful: Technology isn’t just about performance, cost, or scalability. It’s about impact. Impact isn’t measured only in KPIs or deployment success rates. It’s also in the smile of a child when you finally shut your laptop and join her on the floor. Since then, I’ve made it a personal rule: If a system can be automated, I automate it. If a process can be simplified, I simplify it. Not just for efficiency — but to win back time for the things that matter. Because the ultimate goal of technology is not just to make businesses faster, but to make lives better. If you work in tech, ask yourself today: Is your code helping you connect with people who matter?

9

Every engineering leader says they care about velocity. But velocity doesn’t die in the IDE. It dies in: Review queues Approval loops Dependencies across 3 other teams The fastest teams don’t code faster. They unblock faster.

1

💡 AI doesn’t just change how code is written – it changes expectations and outcomes. In the final section of Jellyfish’s AI Adoption Guide, we explore how expectations for engineering leaders are evolving as AI becomes embedded across the SDLC. Because the organizations pulling ahead are treating AI as a transformation, not a tooling upgrade. Inside, you’ll explore: - Why AI adoption is quickly becoming a leadership mandate, not an option - How to balance ambitious goals with realistic expectations - What cultural investments are required to sustain long-term gains - How data can be used to align executives and engineering teams The teams that win in the AI era won’t just have better tools – they’ll have leaders who know how to guide change. Download today: https://lnkd.in/dut_6WpQ

20

Your CI/CD bill is not exploding because of “expensive runners,” it is bleeding out through flaky tests, over-parallelized jobs, and slow feedback loops that quietly tax every deploy This piece shows you how to redesign pipelines so you cut compute waste, shrink queues, and speed up developer flow without adding yet another approval gate or YAML religion Full breakdown in the comments. #DevOps #CICD #PlatformEngineering

8

1 Comment

🚀 Weekend Learnings: CrewAI in Action This weekend, I explored CrewAI—a powerful multi-agent orchestration framework. While it's not a replacement for LangGraph, I was genuinely impressed by how easily I could build collaborative agent workflows with clear roles, tools, and tasks. The concept of giving agents "jobs" and letting them interact feels intuitive and productive. For anyone exploring agent-based systems, CrewAI makes multi-agent coordination feel natural. Highly recommend giving it a try. ✅ Next up: experimenting with custom tools and persistent memory! 💬 Curious—what frameworks or tools have you found useful in building intelligent agents? #AI #LLM #CrewAI #LangGraph #MultiAgent #WeekendLearning #OpenSource #AIEngineering #PromptEngineering #LLMTools #LearningInPublic

84

4 Comments

Playing around with numbers arrangement and I like this particular one: "123498765" I laugh it off as "Disruptive Numerical Symmetry" There is ascent : 1 to 4 Descent: 9 to 6 Disruption: 5 I can't say if 9 is the fulcrum.

1

Curious how a great autocomplete experience works under the hood? Check out this deep dive by our teammate Drew Tate on building it for the ES|QL editor in Kibana:

4

Nate asked what makes Editor V4 so much faster. The simple answer is a new, CSS-first framework that produces incredibly clean code. This means faster load times and higher performance scores for everything you build. Want the full story? Watch Roi explain how this gives you a real performance edge.

29

1 Comment

A Layover Turned Into Being Stranded - Quite a Story I was supposed to attend and present a demo at DVCon U.S. 2026 along with my senior colleague. What was meant to be a 1.5-hour layover in Dubai turned into nearly 6 days of being stranded, which unfortunately meant missing my conference! While flying from Dubai to San Francisco, the aircraft I was on entered Iranian airspace and then, after 3–4 hours of cruising, turned back due to the escalating situation in the region. After circling for a while, we landed back in Dubai. What followed was a long night of confusion at the airport before arrangements were made for stranded passengers. To their credit, the airline arranged hotel accommodation and food for everyone, which I’m genuinely grateful for. Soon a routine formed - airline officials would visit the hotel lobby every morning and evening, passengers would line up while flights were manually reshuffled. Seat confirmations were based on priorities like families with children and senior citizens first, and then solo passengers like me! Standing in those queues twice a day, the engineer in me kept wondering if airlines could leverage AI/LLMs during disruptions like this - analyzing passenger constraints and flight availability to suggest rescheduling options faster, reducing both response time and human effort. Of course, easier said than done, as it would require deep integration with airline operational systems and dedicated engineering teams on the go. During those days there were multiple drone strikes reported in the same city where I was staying (fortunately about 25 km away from me). The area around me remained calm, except for the midnight potential strike alerts on the phone! It definitely added an unexpected sense of adventure and naturally kept my family tensed back at home! Finally, on the sixth day, I got my ticket confirmed back to India. Relief! Immigration done. Security cleared. I was sitting at the boarding gate looking at the aircraft right in front of me through the glass wall. Then suddenly everyone’s phones rang - an emergency alert asking people to seek shelter due to a potential missile strike! We were called back inside and waited for about an hour. Thankfully nothing happened, boarding resumed and the flight finally departed. I’m very grateful to my colleagues/friends who constantly checked in on me, tracked the situation, evaluated alternate routes and suggested travel options throughout the week. I missed a conference, the interactions and learning around, and the post-conference exploration plans! But looking back, life seems to have given me quite a timeline of experiences - studying engineering and starting my career in a pre-AI world, surviving a global pandemic and now getting stranded during a geopolitical crisis! Hopefully this was a once-in-a-lifetime experience… but at least I now have a story to tell for the rest of my life :)

100

4 Comments

Tech hiring in January moves quickly. Teams come back with approved headcount and very little patience for half-ready candidates. If you are in tech and thinking about a move in Jan 2026, try this plan instead. 🧭 Audit your last 12 months 🛠 Tighten your technical story 💬 Prep for technical conversations 🔗 Make your work visible 📡 Signal availability early 🗓 Choose quality over volume January favours candidates who are clear, confident, and prepared. What part of your tech profile are you sharpening for 2026? #TechCareers #TechJobs #TechTalent #CareersInTech #JobSearch #Emmbr #recruitmenttips

6

We published our eval of Claude's Sonnet 4.6 and adaptive thinking. Sonnet 4.6 nearly matched Opus 4.6 on our hardest production investigations. That's a meaningful cost-performance shift for running agents in environments like production systems. Production incidents don't tell you upfront where the hard reasoning lives. An investigation that starts routine can turn novel in a single data point. Adaptive thinking (light reasoning for evidence gathering, deep reasoning for correlation and root cause) handles this without having to specifically calibrate it. Link to the full eval from our team in the comments below:

147

4 Comments

Custom Apigee developer portals work. Until you have to keep evolving them. Auth changes. Docs drift. New APIs appear. Ownership never really goes away. Launchpad is a managed developer portal for teams that want stability without constant rebuilds.

1

162M records/second with seq+compression. Parquet who? Reality check: 90% of the time we need the entire record anyway. Parquet optimizes for "I only need columns A, C, and F" — but then your next query needs A, B, G, and K. And the one after that needs everything. So why the columnar complexity? Especially when Apache Parquet for C++ is genuinely garbage. Bloated, slow, memory-hungry mess. Even the ClickHouse team said "nope" and rewrote their own lean implementation. Seq + fast compression gives me: — Zero schema headaches — Predictable performance — The whole record, every time — 4-10x compression ratios anyway (sometimes zstd, sometimes lz4/xz) Parquet gives me months of optimization theater for edge cases that change next quarter. When even ClickHouse rewrites your "standard," maybe it's time to question the standard.

1

Hitting rate limits at scale? Here’s how we handled it. Our product relied on a third party service with strict rate limits. As usage spiked, users got locked out of event chats—unless they manually retried joining, which led to a poor UX and even lost users. I led the redesign: - Used Kafka to decouple and async the flow. - Leveraged Redis for smart retry windows. - Switched to bulk API calls for efficiency. The result: a scalable, fault-tolerant system built with tools our team already trusted. Wrote up the full breakdown here: 🔗 https://lnkd.in/dUh_YFfF Would love to hear how others have tackled similar bottlenecks.

10

Spotify's engineering team built a background coding agent with Claude Agent SDK that handles complex migrations across their massive codebase. Engineers describe what they need in plain text—via Slack, even—and the agent takes it from there, running tests and opening pull requests. They're now merging 650+ agent-generated PRs into production each month. Read more about how they built it, and what else Claude is enabling at Spotify: https://lnkd.in/gVhivXu6

4,365

164 Comments

🔥 New in Episode 3 of Stacked Sessions: Most teams using AI are measuring the wrong things entirely. They're tracking code velocity while missing what's actually happening to their value delivery. Danny Presten reveals why traditional value stream maps became obsolete the moment AI entered your codebase - and shares the blueprint his teams are using right now. 🎙️ Listen here: https://hubs.ly/Q03VLWwp0 What metric are you tracking that might be lying to you? #EngineeringLeadership #ValueStreams #AI #StackedSessions

13

1 Comment

Interesting article from the VP of Ads AI Core Infra. Worth a read to see how teams like mine are working through putting AI into production at scale.

8

LiteLLM is an open-source proxy that connects your app to LLMs through one unified interface. Paired with Redis, it gives AI and ML teams a simple yet powerful way to unify access to LLMs, accelerate response times, and make AI apps real-time. Redis handles performance, memory, and data coordination that modern AI apps demand, while LiteLLM handles abstraction and routing. Our AI Product Marketing Manager Rini Vasan dives into: ▶️ What LiteLLM is ▶️ Why it works well with Redis ▶️ How LiteLLM and Redis work together Learn how to scale your LLM gateway with LiteLLM & Redis: https://lnkd.in/gqsTebF7

32