Learning in public · current status: foundations

My ML learning journal

I am a senior mobile engineer studying machine learning in public. This is a learning log, not a professional credential page: the roadmap, reading, experiments, and honest gaps.

See the roadmapRead the learning log

Useful mobile context, openly incomplete ML experience

I bring more than a decade of production mobile experience across Android, Kotlin Multiplatform, streaming products, and native performance work. That gives me a practical lens on latency, battery, privacy, and reliable delivery.

It does not make me an ML engineer. Training, fine-tuning, mathematics, data work, and MLOps remain learning goals. I label them that way here.

I have shipped production AI features on my own products. The tool calling concierge on paglipat.com (case study) is the deep one. The floating chat on this site is a live demo. That evidence informs the applied AI capability. It is not client AI work.

What I am learning, in order

Status markers describe current progress rather than aspiration.

  1. Phase 0

    In progress

    Foundations

    • Linear algebra, calculus, and probability refresher
    • Python with NumPy, pandas, and matplotlib
    • Reproducible notebooks and local setup

  2. Phase 1

    Next

    Classical machine learning

    • Regression, classification, and clustering
    • Feature engineering, cross-validation, and evaluation
    • End-to-end work on a real dataset

  3. Phase 2

    Next

    Deep learning

    • PyTorch tensors, autograd, and training loops
    • Vision and sequence-model foundations
    • Practical coursework and small projects

  4. Phase 3

    Next

    LLMs and applied AI

    • Transformers, attention, and tokenisation
    • Grounding, tool use, and model APIs
    • Prompt, tool, cost, and evaluation tests

  5. Phase 4

    Next

    On-device and edge ML

    • TensorFlow Lite and Core ML integration
    • Inference across Android, iOS, Rust, and KMP
    • Quantisation, compression, and latency budgets

  6. Phase 5

    Next

    MLOps and shipping

    • Serving, monitoring, and drift detection
    • Experiment tracking and ML delivery pipelines
    • Shipping one end-to-end ML feature

What I have written so far

The archive stays small until there is real work to report.

Resources in the plan

  • Mathematics for Machine Learning

    Deisenroth, Faisal, Ong · Book

  • Hands-On Machine Learning

    Aurélien Géron · Book

  • Deep Learning Specialization

    Andrew Ng, Coursera · Course

  • Practical Deep Learning for Coders

    Fast.ai · Course

  • The Transformers Library

    Hugging Face · Documentation

  • Designing Machine Learning Systems

    Chip Huyen · Book

Follow the honest version

I publish what I learn, build, and fail to understand on the first pass.

View journal articles

What this page does and does not claim

Is David already working as an ML engineer?

No. David is a senior mobile engineer studying the deeper ML stack in public. Independent applied-AI products are not presented as professional model-training or MLOps experience.

Why connect mobile engineering and ML?

The mobile background creates useful questions about latency, privacy, battery, model size, and shipping reliable on-device features. It does not replace the mathematics, data work, and training knowledge still being learned.

Do you offer AI/ML mentoring?

No. Paid mentoring remains focused on established Android, Kotlin, KMP, architecture, leadership, and career experience.