Building A TA Portfolio

What A Lead Sees In Ninety Seconds

Whoever reviews your portfolio is doing it between meetings, on the second monitor, with forty other applicants in the queue. Here's the actual scan order I have seen leads use again and again:

Do the thumbnails suggest tools and systems, or just renders? A node graph, a UI screenshot, or a before/after GIF signals "TA" instantly. Ten sculpts signal "artist who applied to the wrong posting."
Is there a breakdown? They click your best-looking piece and look for the how. No breakdown = no evidence = polite rejection.
Is there code they can read? A GitHub link with one decent README outranks a paragraph claiming Python proficiency.
Does anything show shipping instinct? Constraints, budgets, performance numbers, artist feedback - anything proving you know work has users.

Optimize for that scan. Everything else on this page is detail.

The Starter Formula: One Tool, One Shader, One Pipeline

Three pieces. That's a complete junior TA portfolio. More is fine, but these three cover the whole territory a hiring lead needs to see:

Piece	What it proves	Examples (steal these)
One tool	You can turn someone's annoyance into a button	A batch renamer with a UI - an export validator - an auto-LOD script - rebuild one of the browser tools inside a DCC
One shader	You understand the GPU side of the craft	A dissolve effect - stylized water - triplanar blend material - a master material with sensible parameters - see Shaders & HLSL
One pipeline	You can make two programs cooperate	Blender -> Unreal batch exporter with correct scale/axes - a texture-processing watch folder - Houdini HDA driving assets into an engine

Every piece must be finished - ugly-but-working beats gorgeous-but-abandoned. Finished means: someone else could use it, and there's a breakdown explaining it.

Anatomy Of A Breakdown

The breakdown is the portfolio. Structure each one exactly like a good postmortem:

The problem, in one sentence. "Artists were spending ~15 minutes per prop setting up export hierarchies by hand."
Constraints. Engine version, team size, why the obvious solution wasn't available. Constraints are what make your choices interesting.
The approach - including the dead ends. The path that didn't work is often the strongest part. It proves the final design was chosen, not stumbled into.
The result, with a number if humanly possible. Time saved, tris reduced, draw calls cut, clicks eliminated. Even rough numbers ("about 15 minutes -> about 30 seconds") transform a claim into evidence.
What you'd do differently. One honest paragraph. Self-awareness reads as seniority.

From the trenches

A candidate once ended a shader breakdown with: "This version is my third attempt. The first died because I didn't understand depth sorting, and I've included it because the mistake taught me more than the fix." That single sentence got them the interview. Every reviewer in the room had made the same mistake once.

Showing Code Without Apologizing For It

Put it on GitHub, public, with a README that has a screenshot, a one-paragraph what-and-why, and install steps. The README is read ten times more than the code.
Pin your best three repos. Delete or hide the tutorial-follow-along ones; they dilute.
Comment the why, not the what. Reviewers open one or two files. Docstrings at the top of each and one clean function are worth more than exhaustive coverage.
Don't hide student code. Nobody expects production polish from a junior - they expect naming that isn't final_v2_REAL.py and functions shorter than a screen.

Hosting & Format

The boring, correct answer: a simple website or ArtStation page for visuals, GitHub for code, and a 60-90 second video per major piece. Video matters more for TAs than for artists - tools are interactions, and a GIF of the tool doing its job communicates more than any still. Keep videos silent-friendly (captions or on-screen text); reviewers watch muted.

One page, no login walls, no "portfolio available on request." If a lead can't reach your work in one click from your resume, the work doesn't exist.

The Classic Mistakes

Mistake	Why it hurts
Twenty pieces, no breakdowns	Reads as artist portfolio; TA signal is zero
"Proficient in Python, Maya, Houdini, Unreal, Unity, Blender..."	Claims without artifacts read as inverse credibility
Only tutorial results	Leads recognize every popular tutorial's output on sight. Twist it: change the goal, break it, extend it, and say so
Hiding the failures	Dead ends are your best material - they show reasoning
No video of tools	A tool that's only screenshots may as well not run
Waiting until it's "ready"	Ship at three pieces. Iterate in public. The portfolio is a pipeline, not a monument