Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
The state of NLP in production 🥽
Search
Abdur-Rahmaan Janhangeer
August 27, 2023
Programming
0
140
The state of NLP in production 🥽
NLP in production vs real life
Abdur-Rahmaan Janhangeer
August 27, 2023
Tweet
Share
More Decks by Abdur-Rahmaan Janhangeer
See All by Abdur-Rahmaan Janhangeer
Building AI Agents with Python: A Deep Dive
osdotsystem
0
30
Extending Flask using the Flask Plugins API
osdotsystem
0
120
PEPs that hit the front page
osdotsystem
0
96
libSQL: Taking Sqlite To The Moon
osdotsystem
0
210
Boosting Python With Rust 🚀
osdotsystem
0
210
Flet: Flutter in Python
osdotsystem
0
450
SQLite Internals: How The World's Most Used Database Works
osdotsystem
2
3.7k
Fast Flask Dev For Big Codebases
osdotsystem
0
230
Python Bytecode or How Python Operates
osdotsystem
0
310
Other Decks in Programming
See All in Programming
Pull-Requestの内容を1クリックで動作確認可能にするワークフロー
natmark
2
520
Catch Up: Go Style Guide Update
andpad
0
230
Domain-centric? Why Hexagonal, Onion, and Clean Architecture Are Answers to the Wrong Question
olivergierke
3
910
Server Side Kotlin Meetup vol.16: 内部動作を理解して ハイパフォーマンスなサーバサイド Kotlin アプリケーションを書こう
ternbusty
3
210
AI Agent 時代的開發者生存指南
eddie
3
1.8k
オープンソースソフトウェアへの解像度🔬
utam0k
16
3k
CSC305 Lecture 05
javiergs
PRO
0
220
テーブル定義書の構造化抽出して、生成AIでDWH分析を試してみた / devio2025tokyo
kasacchiful
0
170
いま中途半端なSwift 6対応をするより、Default ActorやApproachable Concurrencyを有効にしてからでいいんじゃない?
yimajo
2
440
Foundation Modelsを実装日本語学習アプリを作ってみた!
hypebeans
0
120
登壇は dynamic! な営みである / speech is dynamic
da1chi
0
340
アメ車でサンノゼを走ってきたよ!
s_shimotori
0
230
Featured
See All Featured
ReactJS: Keep Simple. Everything can be a component!
pedronauck
667
120k
Creating an realtime collaboration tool: Agile Flush - .NET Oxford
marcduiker
34
2.3k
Site-Speed That Sticks
csswizardry
13
910
The Art of Delivering Value - GDevCon NA Keynote
reverentgeek
16
1.7k
RailsConf 2023
tenderlove
30
1.3k
JavaScript: Past, Present, and Future - NDC Porto 2020
reverentgeek
52
5.6k
What’s in a name? Adding method to the madness
productmarketing
PRO
24
3.7k
Put a Button on it: Removing Barriers to Going Fast.
kastner
60
4k
実際に使うSQLの書き方 徹底解説 / pgcon21j-tutorial
soudai
PRO
190
55k
Improving Core Web Vitals using Speculation Rules API
sergeychernyshev
21
1.2k
Unsuck your backbone
ammeep
671
58k
Responsive Adventures: Dirty Tricks From The Dark Corners of Front-End
smashingmag
253
22k
Transcript
The state of NLP in production
None
Python Mauritius Usergroup site fb linkedin mailing list 3
url pymug.com site 4
About me compileralchemy.com 5
slides 6
The state of NLP in production 7
Hardest part of a real-world project 8
? 9
Is it cooking up an awesome model? 10
No, the world is more complex than this 11
Elements of an NLP project 12
NLP project gather data clean store train use model retrain
model 13
gather data 14
Toy project use curated data set quick extraction 15
Real project a lot of data needed data corresponds to
business case. data probably does not exist speed of data gathering find ingenious / better ways of getting data automate collection 16
clean/preprocess data 17
Toy project use an existing parser / curator e.g. NLTK
existing options 18
Real project use a parser intended for it, several custom
steps parallel processing of data 19
store data 20
Toy project laptop 21
Real project cloud database hot / cold data TTL 22
training 23
Toy project use laptop / external GPU 24
Real project on cloud training on cloud knowledge cross-cloud skills
fault tolerance 25
use model 26
Toy project local website / code 27
Real project continuation of pipeline web service architecture devops /
deploy 28
retraining 29
Toy project euhh this even exists???? 30
Real project learn cloud offerings for continuous learning ways to
retrain / fine tune 31
It's more than serving a model 32
Operation model 33
[ pipeline ] data collection --- process --- train -<-
| | --------------------------- model ^ | | | | --->--- V web service [pod] [pod] --- happy user | -> users service [pod] [pod] | -> db service [pod] 34
skills chart 35
skills --------------- --------------- | | | | | backend |
| devops | | | | | --------------- --------------- --------------- --------------- | | | | | backend | | data eng | | | | | --------------- --------------- 36
skills --------------- --------------- | | | | | backend |
| devops | | | | | --------------- --------------- web service deploy --------------- --------------- | | | | | ml | | data eng | | | | | --------------- --------------- models pipelining 37
code blueprint [ architecture repos ] [ pipeline repos ]
[ ml repos ] [ backend repos ] 38
Tools 39
Pandas Good queries Much resources Read SQL 40
Dask Good for it's purpose: Parallelize tasks Poor docs 41
Polars Awesome parallelizations Great docs 42
NLTK use spacy if possible 43
Notebooks great for cloud used in production on the cloud
44
Advice to research / scientists folks keep everything clean people
will come after you always in hurry / messy / i'll clean it later mood good practices? is this phrase in the korean dictionary? 45
General advices have great docs good onboarding have great standards
46
Keep learning! 47