The latest articles on DEV Community by Rallypi (@rallypi). https://dev.to/rallypi https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3862454%2Fcce4bc65-4007-495c-b02a-d8a01dd0e78b.png https://dev.to/rallypi en Rallypi Sun, 05 Apr 2026 15:36:03 +0000 https://dev.to/rallypi/i-built-an-ai-to-analyze-stocks-after-losing-money-trading-33cp https://dev.to/rallypi/i-built-an-ai-to-analyze-stocks-after-losing-money-trading-33cp <ol> <li>Why I built this</li> </ol> <p>I kept losing money trading U.S. stocks, and realized most of my time was spent reading news and trying to interpret it.</p> <p>After a while, it became repetitive and exhausting.</p> <p>So I built a small tool for myself.</p> <ol> <li>What it does</li> </ol> <p>It takes in news and SEC filings, and turns them into structured daily summaries.</p> <p>Instead of reading dozens of articles, I can quickly see:</p> <ul> <li>what moved the market</li> <li>key takeaways per ticker</li> <li>overall market flow after the close</li> </ul> <p>3.Key features</p> <ul> <li>Daily market close reports </li> <li>On-demand per-ticker analysis (cached after first run) </li> <li>ETF composition and trend insights </li> </ul> <ol> <li>Challenges</li> </ol> <p>One of the hardest parts was balancing cost vs performance.</p> <p>Initially, I wanted real-time generation, but LLM + data API costs were too high.</p> <p>So I switched to:<br> → generate once after market close<br><br> → cache results for fast access </p> <p>This reduced costs significantly.</p> <ol> <li>What I learned</li> </ol> <ul> <li>AI feels more like amplification than replacement </li> <li>Small problems can take an entire day to solve </li> <li>Infrastructure (AWS, CI/CD) was harder than expected </li> </ul> <ol> <li>Open question</li> </ol> <p>Even with sufficient ECS + RDS resources, the first request feels slow.</p> <p>Subsequent requests are fast (due to caching), but initial load is delayed.</p> <p>My guess is:<br> → cache miss → generation pipeline → LLM latency </p> <p>Curious if anyone has dealt with similar issues.</p> <ol> <li>Link</li> </ol> <p>If you're curious, here's what I built:<br> <a href="https://rallypi.com" rel="noopener noreferrer">https://rallypi.com</a></p> ai webdev startup aws