Akhil GhoshAG

questions • projects • loose notes

All models are wrong; some are useful. George Box.
Taped polaroid of a Chicago Bears route tree
CURRENT LOCATION

Chicago, IL

41.8781° N

87.6298° W

Hand-drawn Chicago skyline
A stack of books beside a dark ceramic mug
notesmemorieswritingsystemsarchiveabout
field notes / essays / project logs

Writing

A small archive of things Akhil has been working through: sports analytics, data systems, writing tools, career notes, and the occasional stubborn personal essay.

latest note

Caleb Williams Looks Bad Until You Look Closer

A pizza chart of Caleb's 2025 percentile ranks tells a more complicated story than the raw stats let on. I built the viz with Python and mplsoccer, and the numbers were surprising.

Mar 10, 20264 min read
02

Building a Player Tracking Pipeline for NFL All-22 Film

Zero-shot player detection on NFL coaching footage using SAM3 and SAM2.1. What I built, what broke, and what the output looks like.

Mar 4, 20265 min read
03

Breaking Down the Bears' 2025 Passing Game by Route

A route-by-route look at Caleb Williams' EPA efficiency and how Chicago's receivers were used in the 2025 season.

Feb 26, 20267 min read
04

Data Engineering Runs on Judgment Calls

Two projects, two completely different domains, same lesson. The actual work was sitting in ambiguity and making decisions I could defend. The SQL took care of itself.

Feb 20, 20266 min read
05

Why Pass Rush Win Rate Lies to You

PRWR is one of the most cited stats in NFL analytics, and one of the most misused. Here's what it actually measures, where it falls short, and why the 2025 season is a good case study.

Jan 31, 20263 min read
connect
ResumeLinkedInGitHubEmail

listening mode

curiosity routing
Taped polaroid of the Loyola wolf sculpture
Etymology corner card defining curious
Taped polaroid of a runner in a Loyola uniform
Note to self: Depth is just attention with better taste.