Working for clients and having the clients provide the software / tools / infra is so inefficient it borders on comical. Here is what we have been given for an "AI" contract at an international, 100k person company: 1. laptops from 2015 with no batteries (cant make this up) 2. VSCode is the only allowable IDE, no language run times or servers available on your local machine (aka no syntax checking, or debugging available locally) 1. no VScode plugins 2. took 3 weeks for IT to do the vscode install ticket btw 3. All files have to be FTP to a server to be executed. 1. that server is used by 200+ analysts a day: there is no resource queue, first come first serve. If someone decides to train BERT on CPU they can block the entire platform for DAYS 2. this server has no python package management: sysadmins keep installing everything to the same base python 3.6; dependency graph over writes are almost a weekly occurrence 4. VPN that is down maybe 10% of the workday 5. kafka esque data access controls: 1. i got cleared to view their PII on server "foo" (after 1 month of signing pdfs) 2. PII data is also uploaded to HDFS/spark, same data, same tables: "select * from foo" 3. not allowed to read the data on spark because its PII 6. 1 gpu (again, hundreds of people on this AI contract) 7. Not a single API for this "AI" 1. predictions will be emailed in a csv every day. 2. no crontabbing the scripts either: thats a "security risk": someone has to ssh in and run predict.py TWICE A DAY even on weekends. (obviously first thing i did was just have a loop.py that fires every 8 hours and runs main.py) How does anyone expect anything productive to get done? Is this the norm for contracting "AI"? The client is paying millions of dollars for this and they're giving us mud and sticks to work with, no option for us to use our own tools or any cloud tools. don't even get me started on the meetings... #data #dataanalytics #datascience #machinelearning #clients #client #devops #infrastructure
This sounds like a bank. They really like doing AI work š
This has been my experience as well. Consulting in analytics is horrible IMO
World Conflicts
Yesterday
584
Israeli precision-guided munition likely killed group of children playing foosball in Gaza, weapons experts say
World Conflicts
Yesterday
501
Is "From the River to the Sea" So Wrong?
India
Yesterday
635
'Hindutva': The Radical Hindu Ideology That Seeks to 'Push Christianity Out of Indiaā
Tech Industry
Yesterday
1535
Do people underestimate E6 role at meta?
Personal Finance
Yesterday
1110
Thank you AAPL and NVDA
You deserve better. Find a job that lets you work with people who aren't clowns
Take me with you to the FAANG promised land where when I say ātest suiteā people donāt look at me funny.