# Mission: ML Pipeline (STOR 323) — for Sage

## Why
Sage is a college student taking STOR 323 (the same course this Lecture 3 deck is from). She understands high-school math but hasn't done formal stats/ML before. The goal here isn't procedural fluency for its own sake — it's to **challenge her to think about the big concepts**, using the intellectual tools she already trusts: probability (an existing interest of hers), visual/logical reasoning, and zen and Hindu philosophy. The bet is that ML's core ideas — uncertainty, estimation vs. truth, cause and action, the ethics of a rule — aren't new to her at all; they're old philosophical questions wearing new notation. The job is to make those connections explicit and load-bearing, not decorative.

## Who this is for
Sage. Not a generic student — lessons should speak to her directly, assume the philosophical vocabulary is already meaningful to her, and go deep rather than sprinkle a surface reference and move on.

## Success looks like
- Can explain core ML ideas (fitted model vs. true model, training vs. deployment, thresholds/policy) *and* articulate the philosophical parallel that made each one click.
- Treats "we can never know f, only estimate f̂" as an old, familiar epistemic stance — not a new and uncomfortable one.
- Can hold both registers at once in an exam or problem set: the formal math when required, the intuition underneath it when reasoning about a new situation.
- Comfortable working STOR 323 problem sets and exam questions as they're released, same as any student in the course.

## Constraints
- High-school math baseline — define statistical/linear-algebra notation the first time it appears, same care as any first-time learner, just don't over-explain the philosophy she already has fluency in.
- Sessions keyed to the same lecture deck sequence as the course delivers it.

## Philosophical approach
**Deep integration, not seasoning.** Every lesson should build the ML concept and the philosophical parallel together, as two views of the same idea, with real primary-source grounding on the philosophy side (not vague "this is very zen" gestures). Candidate throughlines for this deck already identified:
- **Samskara** (Yoga Sutras) — mental impressions formed by accumulated experience, which then condition future perception — as the lived analogue of a fitted model f̂_D formed from data D.
- **The finger pointing at the moon** (Zen) / the **Two Truths** doctrine — f̂ (the estimate) is never f (the true, unknown thing); the map is not the territory.
- **Nishkama Karma**, Bhagavad Gita 2.47 ("you have a right to action, not to the fruits of action") — the deployed policy δ as a question of right action under uncertainty, distinct from the model/knowledge itself.
- **Anicca** (impermanence) / epistemic humility — why probability, not certainty, is the honest description of the world.

## Out of scope
- Treating the philosophy as flavor text or a hook to "make math fun" — if a parallel doesn't hold up rigorously on both sides, cut it rather than force it.
- Comparative religion for its own sake — philosophy is in service of the ML understanding here, not a second subject.