My Dashboard

November 2022 launch, RNNs vs Transformers, tokens, embeddings, and what each letter means.

CUDA cores, Tensor cores, memory hierarchy, and NVLink — what’s actually inside a GPU.

Worked calculation for Llama 70B: weights, quantization, KV cache, and runtime overhead.

One AWS door to Claude, Llama, Nova, and more — Converse vs InvokeModel, billing, architecture, and production metrics.

An always-on SRE teammate: Agent Spaces, skills and memories, incident investigation, and agent-second pricing.

GPU Operator, Device Plugin, MIG, scheduling, and your first nvidia.com/gpu pod on the cluster.

Function calling, tools, Chain of Thought, ReAct, and no code / low code / full code frameworks.

Trigger on a pull request, fetch diffs, ask OpenAI for a DevOps review, and post it back on GitHub.

An AI agent that uses Linux tools to collect evidence, then explains why the server is slow.

Read Apache logs with LangChain and OpenAI, then get a human-friendly explanation of the errors.

Why AI apps need a standard tool protocol — Host, Client, Server, JSON-RPC, stdio, and Streamable HTTP.

Ingest docs, chunk, embed, retrieve Top-K context, and generate answers grounded in your knowledge base.

Base models, SFT, LoRA, QLoRA, rank, adapters, and how fine-tuning steers model behavior.

Fine-tune with lakhera2023/dockerNLcommands-sft-unsloth — natural language Docker requests to real CLI commands.

Question, hint, Judge0 editor, and an AI coach that only gives nudges.

Verbal SRE / Cloud scenarios by company — structure first, coach nudges only.