Video logs

𝗔𝗜 𝗔𝗣𝗜𝘀 𝘃𝘀 𝗦𝗲𝗹𝗳-𝗛𝗼𝘀𝘁𝗶𝗻𝗴: 𝗪𝗵𝗶𝗰𝗵 𝗢𝗻𝗲 𝗠𝗮𝗸𝗲𝘀 𝗦𝗲𝗻𝘀𝗲? AI APIs may seem expensive as usage grows. Self-hosting may look like a cheaper alternative, but the decision involves more than API and GPU pricing. 𝗔𝗜 𝗔𝗣𝗜𝘀 𝗮𝗿𝗲 𝗼𝗳𝘁𝗲𝗻 𝗮 𝗴𝗼𝗼𝗱 𝗳𝗶𝘁 𝘄𝗵𝗲𝗻: • You are building an MVP • You need fast development • Usage is low or unpredictable • You want to avoid infrastructure management 𝗦𝗲𝗹𝗳-𝗵𝗼𝘀𝘁𝗶𝗻𝗴 𝗺𝗮𝘆 𝗯𝗲 𝘄𝗼𝗿𝘁𝗵 𝗰𝗼𝗻𝘀𝗶𝗱𝗲𝗿𝗶𝗻𝗴 𝘄𝗵𝗲𝗻: • AI…

When your application starts slowing down, there are usually two options. • 𝗨𝗽𝗴𝗿𝗮𝗱𝗲 𝘆𝗼𝘂𝗿 𝗲𝘅𝗶𝘀𝘁𝗶𝗻𝗴 𝘀𝗲𝗿𝘃𝗲𝗿 with more CPU and RAM. This is vertical scaling. • 𝗔𝗱𝗱 𝗺𝗼𝗿𝗲 𝘀𝗲𝗿𝘃𝗲𝗿𝘀 and distribute traffic between them. This is horizontal scaling. 𝗔 𝘀𝗶𝗺𝗽𝗹𝗲 𝗲𝘅𝗮𝗺𝗽𝗹𝗲: An e-commerce website starts getting more visitors. For steady growth, upgrading one server is often the fastest and simplest solution. If traffic suddenly spikes during a big sale, adding more servers i…

The case: "Show a user's transaction history." Team 1 builds a banking app. Team 2 builds a social media app. Same screens. Same API. Same feature. Completely different architecture. That's because of two things every system needs. 𝗙𝘂𝗻𝗰𝘁𝗶𝗼𝗻𝗮𝗹 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝗺𝗲𝗻𝘁𝘀 tell you what the system should do. Show the history. Display the numbers. Load the screen. 𝗡𝗼𝗻 𝗳𝘂𝗻𝗰𝘁𝗶𝗼𝗻𝗮𝗹 𝗿𝗲𝗾𝘂𝗶𝗿𝗲𝗺𝗲𝗻𝘁𝘀 tell you how it should behave. How fast. How accurate. How available. For the bank, a sta…

While working on an AI bot, I noticed that some customer questions were still being missed, even though semantic search was performing well. The solution was surprisingly simple. I combined semantic search with text-based similarity search. Result: • Better answer coverage • Fewer missed queries • More reliable responses • Better customer experience A good reminder that improving AI systems is not always about using bigger models. Sometimes it's about combining the right techniques. 𝘏𝘢𝘷𝘦 𝘺𝘰𝘶 𝘶𝘴𝘦𝘥…