Built by
Akshay Joshi
Founding Applied AI Scientist · Munich
This platform is grounded in Akshay's research and engineering background across agentic AI, multimodal vision-language models, semantic search, and production systems.
At Blinkin GmbH, he leads the Agentic AI and Multimodal Vision Language Model R&D team and has shipped multimodal agentic LLM pipelines for enterprise analytics, automation, discovery, multimodal search, and video understanding.
Blinkin GmbH
Leads the Agentic AI and Multimodal Vision Language Model R&D team. Recent work includes a 40B sparse MoE VLM and a Qwen2-VL 72B video understanding system for multimodal search and RAG.
Google Research + DFKI
Worked on DiagramNet, self-supervised multimodal learning, semantic search, and explainability.
AMD R&D
Built platform security processor firmware and Ryzen Master SDK tooling. That systems background shows up in the platform's focus on latency, implementation detail, and reliability.
Why this platform looks the way it does
- Trade-offs, assumptions, and failure modes matter more than memorized definitions.
- Retrieval, ranking, multimodal reasoning, and serving constraints are treated as core ML topics rather than side notes.
- Code implementations are expected to be production-minded: shapes, vectorization, numerical stability, and system limits.
- Interview preparation is organized around the kinds of questions strong ML researchers and infrastructure engineers actually get asked.