Slides: What Lies Beneath the API — Benjamin Cowen, Modal

Source Video

What Lies Beneath the API — Benjamin Cowen, Modal

Relationship To World's Fair 2026

These slides are extracted from a public AI Engineer YouTube video connected to World's Fair 2026. Speaker-matched clips are supporting context unless later confirmed as exact session recordings; official livestream recordings are day-level/event-level source material.

Related Scheduled Sessions

Extracted Slides

slide-001.jpg

OCR text:

i

, &§

4 aa? a *

e “ < wa

slide-002.jpg

OCR text:

Panes

* 5

bod

What Lies Beneath the API:

When you should fine-tune your own model

Benjamin Cowen, PhD

ae —

; a

as °

“ Google DeepMind

en

slide-003.jpg

OCR text:

Modal 101 Functions Sandboxes

Pon a Storage Application State Led rere clare

. * As a general purpose platform, we're seeing

ra é some interesting patterns.

sre

(OOS Ceres ane

Capacity Orchestration

¥', Modal

. ee

+ #3 Braintrust €} WorkOS OpenAl

slide-004.jpg

OCR text:

ge TA ™~ :

—— 7}

) —

I

2

id

‘ -_

EUROPE

o. ’

slide-005.jpg

OCR text:

Works for

© Get started fast

« —Noinfrastructure overhead

. ® Very good models!

oe Use a frontier API

* va Move fast Snip semetn. aq When you might want to shift

Sd n i bd e You need performance control at scale

a

Cost, throughput, latency, custom metric

© You need model differentiation

© You're hitting rate mits

7 ”

a | Al Engineer |

SU ela

slide-006.jpg

OCR text:

SFT in 300 lines

st of code

Ea bs

* a

ior diel Once you have your data in place, open source

libraries and serverless infra make parallel

hyperparameter sweeps trivial.

laltd Space ea Lonel Es raD rere tt MEO LM ba le-¢ Mss Cran Sh eet one SLES

O6_ap4_ana_m “yalo-finetu re _yolo py

»

re)

S Z | Al Engineer |

aU e) a3

at

slide-007.jpg

OCR text:

GRPO (RL) in 300

a lines of code.

Mare

7 ion ial ’ Unified GPU + Sandbox APIs makes large scale

eee

Perea eee rt ale

, Engineering the future of Al

Slide-Derived Subjects To Review

Subject extraction uses video title, related session titles/descriptions, transcript context, and OCR text when available. OCR is best-effort and should be reviewed against the embedded slide images.