Razvan "Val" Stroe

I built a COMPLETE marketing platform on AWS that handles 10,000+ users and costs $0 to run. AWS Free Tier goes WAY beyond just EC2. A few years ago I launched a fully functional serverless marketing platform that cost me ZERO dollars to run (yes, $0.00!) using AWS Free Tier services: • Lambda for backend processing (1M free requests/month) • API Gateway for RESTful APIs (1M free calls/month) • DynamoDB for database (25GB free storage) • S3 for static website hosting (5GB free storage) • CloudFront for global content delivery • Cognito for user authentication • EventBridge for event-driven workflows No servers to manage. No capacity planning. No upfront costs. Just code and business logic. The serverless paradigm isn't just cost-effective—it's transformative for developers and businesses alike. You focus on building features while AWS handles the infrastructure scaling. Don't let budget constraints stop your cloud journey. The AWS Free Tier is much more generous than most realize! What serverless project are you building next? Drop it in the comments! 👇 • • • If you found this useful.. 🔔 Follow me (Lefteris Karageorgiou) for more Cloud insights ♻️ Share so others can learn as well

95

13 Comments

Tax is a special domain: it requires 100% correctness (you don’t want to compute someone’s taxes incorrectly)! Because of that, we’ve had to build lots of custom infrastructure to deploy agents at Column Tax. We wrote a post detailing the architecture of and how we successfully built and deployed Iris, an AI tax development agent, backed by a custom orchestration system, for our very specialized domain-specific coding task. It’s working: we’re now able to build our tax engine much faster than traditional methods. Check out the blog post in the comments below where we expose the specifics of how we built & deployed this AI agent in our 100% accuracy domain:

192

25 Comments

This is going to be a collection of pent-up observations. I've spent the last month+ talking to a ton of people from the past 30 years of building software, writing code (or more accurately, using AI to generate code) and doing some fractional advisory work with startups. A few highlights: Claude Code: - `Claude usage limit reached. Your limit will reset at 9pm (America/Chicago).` - This was my state yesterday after getting to know Claude Code. Running `claude init` on a new codebase I've never seen before was breathtaking. Having a conversation with Claude about concerns on findings was like condensing a days worth of work into 30 minutes. I can never go back. - But, it turns out Claude has its limits. After about 25 commits in on one problem, I'm going to be diving deep into an MCP client that it just can't seem to figure out how to build. Reconnecting: - I've worked with a lot of very great people over the years. People that have helped me, I've helped, and learned from. - A ton of innovation is going on. There are huge opportunities and gaps in the AI landscape (as I've posted before many of the AI startups are just ignoring the real problems to be solved). Advising: - There is no compression algorithm for experience. The need for strong architectural guidance on solving business problems with organizational alignment and connecting the technical dots with the business is massive. Finding people who can do this is super hard.

61

6 Comments

SaaS scale-ups almost never work with an external dev agency. Why? 👉 Trust. Or rather, the lack of it. But here's the real question: What's riskier - trying an external team or watching your competition launch features while yours sits in the backlog? Sometimes playing it too safe is the riskiest move of all. Here are my 4 tips on how to mitigate the risks: 1️⃣ Start with a simple coffee chat. 15 minutes with their CTO tells you everything. 2️⃣ Test drive with a small project. No big commitment. 3️⃣ Stay in control. Code reviews, stand-ups - treat them as your plug-and-play team. 4️⃣ Check their track record. Good agencies value reputation above all. For more tips on hiring your external partner you can check out this playbook we have prepared: https://lnkd.in/d3HvpSkr If your backlog is growing faster than your team can handle, maybe it’s time to rethink your approach. Let's talk!

104

16 Comments

"But that revenue target is totally fu$%ing impossible!?" I exclaimed a former boss of mine as I showed him the decades long historical trend to no avail. I had spent weeks pouring over the numbers, trained an ML model on the historical data, and had even hired an analyst on fiver to give me a second opinion. But there was no way around it. Accepting this number would doom me and my team to the inevitability of failure. Maybe it was the pressure or maybe it was the lack of sleep but inevitably the last sinnew that attached me to giving a fu$% broke and I found myself able to see the forest from the trees. As Jake McNeill mentions in this Substack article Systems Over Goals, Goals fail because: 1️⃣ They fail because they tie self-worth to results you can’t control. 😵‍💫🎯 2️⃣ They dangle a finish line that doesn’t exist. 🏁➡️🚧 Even when you hit it, your ego (or your boss/investors) moves the goalposts. 🥴⚽️ If I played the “hit the impossible number” game, I was screwed from day one. No amount of decks, dashboards, or ML models were going to save me. Because the game itself was rigged. That’s what Jake is really saying: Goals are a trap. They look like progress, but they’re really just moving finish lines designed to keep you sprinting until you collapse. Systems? Systems don’t give a sh*t about your boss’s fantasy target. ⚡ Systems focus on inputs you can control. 🎯 📈 Systems compound over time. 🔁 🧠 Systems build self-trust — the kind that doesn’t evaporate when your quarterly number doesn’t. 💪 So what did I do after accepting my 🐄 💩 fate? I focused on building systems that I could control and that I knew would compound. At the time we had no lead routing system so it could take north of two weeks to get someone on the calendar. So I built a gangster version of Chilli piper and made sure we religiously practiced a good outbound cadence. Did we hit our goals? Hello no! But my guys still made money and the systems that were put into place are still paying dividends long after my departure.

31

2 Comments