Skip to content
Satya Kishore

04 / Learning

AI engineering journey

A working log, not a course catalog. Statuses are directional: completed foundations, active building, and a lot still marked next. Nothing here is a proficiency score.

  • Completed
  • Building
  • Learning
  • Exploring
  • Next

  1. 01

    Foundations

    The practical base everything else sits on: writing programs, living in a terminal, and keeping work in version control.

    • Python

      Primary language for current AI systems.

      Completed

    • Linux

      Command line as a daily environment.

      Completed

    • Git

      GitHub-centered workflow.

      Completed

    • Programming

      C and Verilog from the ECE/VLSI path, plus software systems.

      Completed

  2. 02

    Machine Learning

    How models are specified, fit, and judged — still being built out, not treated as finished theory.

    • ML fundamentals

      Learning

    • Data

      Learning

    • Models

      Learning

    • Evaluation

      Exploring

  3. 03

    Deep Learning

    Neural nets, training dynamics, and optimization — enough to read papers and debug training, not claimed as research depth.

    • Neural networks

      Learning

    • Training

      Exploring

    • Optimization

      Exploring

  4. 04

    Transformers

    The architecture behind current language models. Working through attention, the stack, and tokenization as implementation knowledge.

    • Attention

      Learning

    • Transformer architecture

      Learning

    • Tokenization

      Learning

  5. 05

    Generative AI

    Using and inspecting large language models: prompting, inference, and how behavior changes with decoding and context.

    • LLMs

      Building

    • Inference

      Ollama locally; Groq for hosted low-latency paths.

      Building

    • Prompting

      Building

    • Model behavior

      Learning

  6. 06

    LLM Engineering

    Grounding models in data. Hybrid retrieval is already in a project; evaluation and learned reranking are still ahead.

    • RAG

      Building

    • Embeddings

      Building

    • Vector databases

      FAISS in the support RAG project.

      Building

    • Hybrid retrieval

      BM25 + vectors fused with RRF.

      Building

    • Reranking

      Exploring

    • Evaluation

      Next

  7. 07

    Agentic AI

    The current center of gravity: tools, planning, verification, and the difference between a loop and a reliable system.

    • Tool calling

      Building

    • Planning

      Building

    • Memory

      Exploring

    • Agents

      Building

    • Workflows

      Exploring

    • Verification

      Core idea in Meera: plan ≠ proof.

      Building

  8. 08

    Production AI

    The next altitude. Some pieces exist in projects (serverless deploy, constrained execution); most of the production discipline is still in front of me.

    • Deployment

      Exploring

    • Observability

      Next

    • Evaluation

      Next

    • Reliability

      Building

    • Optimization

      Exploring