Session recaps, tutorials and guest posts from the AIYatra open-source AI community.https://aiyatra.io/https://aiyatra.io/blog/embeddinggemma-2-a-compact-multimodal-embedding-model/https://aiyatra.io/blog/embeddinggemma-2-a-compact-multimodal-embedding-model/EmbeddingGemma 2 is Google’s latest open embedding model for semantic search and retrieval-augmented generation (RAG). Built on Gemma 4, it maps text, source code, images, video, and audio into a shared 768-dimensional vector space. It is released under the Apache 2.0 license and scales from 270 million to 740 million parameters depending on the modalities you enable.Thu, 08 Oct 2026 00:00:00 GMTGuest postEmbeddingGemmaGemmaMultimodal AIAI EmbeddingsRAGSemantic Searchhttps://aiyatra.io/blog/sft-to-rl/https://aiyatra.io/blog/sft-to-rl/SFT with LoRA, DPO and GRPO on a pretrained model — rewards, preferences and the hard question — better at the task, or just better at the reward?Sat, 12 Sep 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/harnessing-agentic-ai/https://aiyatra.io/blog/harnessing-agentic-ai/Build the loop, the tools and the guardrails — and make your own coding agent that reads, plans, edits, tests and repairs its way to a verified pull request. 100% offline, zero API keys.Sat, 05 Sep 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/pytorch-foundations/https://aiyatra.io/blog/pytorch-foundations/Tensors, autograd and the training loop — the foundations every AI engineer stands on. We rebuilt nn.Linear from scratch.Sat, 22 Aug 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/linear-algebra-to-transformers/https://aiyatra.io/blog/linear-algebra-to-transformers/Vectors to attention — embeddings, Q/K/V projections, LoRA and SVD, the math beneath the models, by hand.Sat, 15 Aug 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/deepseek-v3-from-scratch/https://aiyatra.io/blog/deepseek-v3-from-scratch/Multi-head latent attention, MoE load balancing and RoPE scaling — the DeepSeek-V3 paper, live-coded in PyTorch.Sat, 08 Aug 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/goose-ai-agent-demo/https://aiyatra.io/blog/goose-ai-agent-demo/Providers, context engineering, MCP extensions and recipes — a full end-to-end run of the open-source Goose agent.Sat, 01 Aug 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/speculative-decoding-lab/https://aiyatra.io/blog/speculative-decoding-lab/Draft-verify-accept in 60 lines of PyTorch, then HF assisted generation and llama.cpp — faster inference, proven lossless.Sat, 25 Jul 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/agentic-ai-reading-4/https://aiyatra.io/blog/agentic-ai-reading-4/Book-reading series finale: agentic patterns, discussion and reading together — online, open to everyone.Tue, 14 Jul 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/agentic-ai-reading-3/https://aiyatra.io/blog/agentic-ai-reading-3/Session three of the Hitchhiker's Guide series — agents, tools and workflows, read and debated together.Sun, 05 Jul 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/agentic-ai-reading-2/https://aiyatra.io/blog/agentic-ai-reading-2/Session two of the reading series — continuing the guided tour through agentic AI, online in the evening.Thu, 02 Jul 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/agentic-ai-reading-1/https://aiyatra.io/blog/agentic-ai-reading-1/Where the reading journey began — session one of the Hitchhiker's Guide to Agentic AI, online and free.Tue, 30 Jun 2026 00:00:00 GMTSession recaphttps://aiyatra.io/blog/transformer-paper-to-code/https://aiyatra.io/blog/transformer-paper-to-code/From Attention Is All You Need to running source code — the transformer, line by line, from paper to PyTorch.Sat, 20 Jun 2026 00:00:00 GMTSession recap