The Everything Feed - All Packet Pushers Pods D2DO291: From Politics to Machine Learning and AI Engineering
13 snips
Jan 7, 2026 
Marina Wyss, a Senior Applied Scientist at Twitch, transitions from political science to machine learning, bringing a unique perspective to AI Engineering. She explains the distinctions between AI Engineering and ML Engineering, plus when to use pretrained models versus training from scratch. Wyss highlights the importance of a social science background in framing problems and understanding human-centered ML, while sharing essential resources for aspiring engineers, emphasizing hands-on projects over certificates for career advancement.
AI Snips
Chapters
Books
Transcript
Episode notes
What Sets AI Engineering Apart
- AI Engineering often builds on pre-trained models instead of training from scratch.
- The main distinction is using pretrained models via APIs versus building models end-to-end.
Career Shift From Scratch To Pretrained
- Marina moved from training models from scratch early in her career to using both approaches today.
- She now chooses between custom training or pre-trained models depending on the use case like recommendations or content understanding.
Pick Models By Capability Then Iterate
- Narrow candidate models by modality and capability before prompt testing.
- Then iterate prompts or fine-tune to match your desired output and voice.
