Saran Teja MallelaData engineering, applied AI evals, and RL environments for clinical AI. Notes from the build.https://saran-blog.vercel.app/What a Data Engineer Misses About Transformers (Until You Build One)https://saran-blog.vercel.app/posts/what-a-data-engineer-misses-about-transformers/https://saran-blog.vercel.app/posts/what-a-data-engineer-misses-about-transformers/I move data through pipelines for a living. Then I built the transformer stack from scratch, from autograd to attention. Five places my instincts pointed the wrong way, and three bugs that never crashed.Fri, 03 Jul 2026 22:00:00 GMTWhy I'm Spending 8 Weeks Building Toward Verifiable Clinical AI Environmentshttps://saran-blog.vercel.app/posts/why-8-weeks-verifiable-clinical-ai-environments/https://saran-blog.vercel.app/posts/why-8-weeks-verifiable-clinical-ai-environments/One clinical extraction environment, built once and reused three ways, as an evaluation suite, an RL reward signal, and a governance instrument. The plan, the constraints, and the rule that everything ships on Fridays.Sat, 20 Jun 2026 15:00:00 GMTFour Days of RL: One Gradient, Dressed Four Wayshttps://saran-blog.vercel.app/posts/four-days-of-rl-one-gradient/https://saran-blog.vercel.app/posts/four-days-of-rl-one-gradient/Bellman as a fixed-point computation, watching behavioral cloning fail on schedule, and why REINFORCE is the ancestor of everything in modern post-training. Notes from four days of foundations.Thu, 18 Jun 2026 00:00:00 GMT