Tejas Kumar is an AI engineer at IBM, the author of Fluent React and host of the ConTejas Code podcast. He explains how web and AI systems work through books, code-based articles and conference talks.[1][2]
Start with Fluent React →
X 
GitHub Career
Kumar’s professional biography lists roles across web engineering and developer relations, including work at Spotify, G2i, Vercel and Xata, where he was Director of Developer Relations. He joined IBM through its acquisition of DataStax. His present public work covers language models, retrieval, tool protocols and systems of multiple agents. ConTejas Code extends the teaching format into long conversations with other engineers, while his talks demonstrate systems through code.[1]
Contributions
Fluent React looks beneath the framework’s public functions to explain how a user interface is updated. Its chapters cover how JSX becomes elements, how React compares a new render with the old one, and how reconciliation determines what needs to change. Later material addresses server rendering, concurrency and React Server Components. Understanding those mechanisms helps a developer reason about behavior and performance rather than memorize an API. Kumar’s accompanying book page links talks and a React Fiber article for readers who want to inspect the implementation.[2]
Ideas
His Kolibri article studies an open-weight German and English model from Aleph Alpha. It explains a mixture of experts as a system that activates only some of its available networks for each token, then distinguishes the computation used from the memory needed to hold the full model. The article also examines tokenization, deployment requirements and evaluation limits. Kumar identifies the technical report and model card as his sources and adds his own tokenizer experiment. It is explanatory work about Aleph Alpha’s model, rather than model authorship.[3]
Tools, writing & talks
Writing, research & talks
Start here
Fluent Reacttej.asA book and reading guide explaining React’s rendering and update mechanisms.
Read the book overview and chapter map first. Basic JavaScript and React familiarity will help with the internals; newcomers can follow the linked official React documentation before the deeper chapters.
Author.[2]- Evals in AI: A Deep Dive — Tejas Kumar, IBMyoutube.com
Tejas Kumar builds an eval live: assertions fail on a refund-policy case, then an LLM judge, dataset and human verdicts expose self-preference and bias. He sets an agreement threshold for CI and covers runtime versus pre-deploy checks.
Catalog workOriginal source An explanation of mixture-of-experts architecture, tokenization and practical deployment choices.
Author; analysis of Aleph Alpha’s work.[3]Original sourceLinks to talks, AI engineering articles and the ConTejas Code podcast.
Author, speaker and host.[4]
