Just crossed 5k on Upwork since I started back in November with a $5 project 😂( Back then people laughed at me because of how little I made, but I kept at it. I've played this game before, you start slow and build over time) Now my next goal is to push every single contract towards the 1k mark. I'll still do some contracts for less than 1k just for the experience to use the tools and be able to say that I've used the tool. But besides that if there is no real upside in terms of skill or money I'll just say no. Which reminds me of, I just received another €300 offer but this was straight up not a €300 build more like a €1000 build minimum or even more 😅

The challenge was to build an AI support agent using n8n that can automatically handle incoming customer emails. The agent needs to: • classify emails into categories (setup, pricing, security, HR, escalations, spam, etc.) • generate replies grounded in Nexus Integrations’ documentation • escalate emails to the correct team when required • return responses in structured JSON format The interesting part wasn’t just using an LLM — it was designing the workflow architecture around the AI. Here’s what I built: ⚙️ Email Classification Layer Incoming emails are categorized so the system understands the intent. 🤖 AI Support Agent Generates replies using a controlled knowledge base (pricing, integrations, security policies, escalation rules) to avoid hallucinations. 🚫 Spam & Misdirected Filtering Unrelated emails are filtered before they reach the AI agent. 📦 Structured Output Responses are formatted into JSON so they can be evaluated automatically. 📊 Automated Evaluation Pipeline A separate workflow sends test emails to the agent and scores responses using an LLM judge based on: - category correctness - documentation grounding - correct escalation handling Big learning from this challenge: 👉 Building AI systems is less about prompting and more about designing reliable workflows and guardrails around the model. Handling edge cases, grounding responses in documentation, and designing evaluation loops turned out to be the most important parts. Sharing the workflow architecture below 👇 Curious how others approached the challenge and structured their agents. #n8n #n8nchallenge #automation #aiagents

5 of them became my first paying customers — within 60 minutes. The product isn't even publicly launched yet. Here's what happened. I spent the last year building a desktop voice AI studio as a solo dev. The interesting technical challenge: running 3 different ML inference engines entirely on-device in the same process. Native Rust backend, no Python, no Docker. The fast engine does 6x real-time on Apple Silicon. The multilingual engine handles 23 languages. Voice cloning from a short audio sample. I'm partially dyslexic — been converting text to audio since high school. That personal workaround eventually became a real product. 3 days ago, I emailed ~600 customers from a marketing tool I built back in 2012. Hadn't contacted them in over a decade. Plain text emails, no design, no branding — just personal messages. 10% open rate on a stone-cold list. Some of them actually bought. Revenue before the product even launched publicly. One guy found a PDF from my old product on his hard drive and wrote me a full page about it. 14-year-old business relationships converting into customers today. Tonight it goes live. For anyone here building with AI — has anyone else run ML inference natively on desktop? The symbol collision problem between multiple inference engines was genuinely hard. Curious if others have hit this.

A few days ago I applied to a job on Upwork. I sent over a Loom video of my profile, what I've done and just some proof that I can deliver. They came back a few hours later and booked a meeting. Before the call I received an 18-page PDF on what they wanted. That document scared me to the point where I was second-guessing my skills. But then on the call itself, they revealed that they're looking for an AI agent to reply to customers. Lol. A lot easier than I thought. So here's where things go wrong. I do not diagnose the problem at all. I assume that the problem they present is the problem to solve. So when I agree to the job and that I can deliver the agent, we're basing the entire project on: "what they think the problem is." So I sent over an offer for $1207.75. They try to negotiate it down to $800. I revert back that this covers the entire project. And they counter with adding in more things and accept the deal. A few minutes later I receive the contract. At this point I've locked down the contract and I believe I can complete the job. I have a week to deliver an MVP and I have enough cash to hire help if I can't deliver. But then I receive an invitation to their customer service platform. I start looking around and setting up everything. This is when I start to realize that they're using zero internal automations. And when I start going through their emails I realize that 97% of all tickets use templated replies. The last 3%? Refund/cancellation issues or standard customer service questions. Those 3% are the perfect place to use AI because of the nuances. But the 97%? A simple set of rules and automation handles that. They don't need AI, they need automations that runs based on rules. That will clear all cases within a day. And that's without using AI. Looking at this, I could have figured this out if I had asked them about their setup in the call. I would have offered a simpler solution instead of a complex one without even thinking about it. The lesson is: ask more questions, dig for the problem.

Hello guys, just a few weeks into this AI world and I have already been able to create something sought for. Was on a call the other day and they said they has a problem: When a CV arrives in their company email, they want an instant follow-up to tell the people that their CV is being evaluated. So I got down to it and built it for them. It's simple enough but the fact that it solves a real problem is the most important. What do you think? Any improvements? PS: Creating the IMAP and SMPT email credentials was such a pain😆 Anybody relates?