DataEngBytes 2023 - BRI-T1-02 - Mikiko Bazeley

on YouTube ↗39m

" summary

Argues that the full-stack data scientist remains essential because MLOps is fragmented and one-size-fits-all stacks fail in practice, with concrete patterns shown: batch via GCP + Spanner for offline inference, containerized model service for live predict, and real-time streaming in-game recommendations at Riot. Calls out tools people actually use—Kubernetes, Docker, CUDA, vector DBs, embeddings, GPT-3.5/ChatGPT—and notes practical adoption pain: many vendor options, integration work, and a 170-slide deck of real-world lessons.

AI-written summary · about

" chapters (7)

1. 0:00Start
2. 4:09Why Does The Question Even Matter?
3. 4:43Everything is Becoming ML
4. 8:03Here Comes Prompt Engineering
5. 14:41Solving The 3V's of Production Data Science
6. 23:55Key Snapshots Of The LLM Convo
7. 28:19Current Challenges of Running LLMS

" related talks

• 30mDataEngBytes 2023 - MEL-T2-01 - Mikiko BazeleyMikiko Bazeley · DataEngBytes Melbourne 2023
• 28mDataEngBytes 2023 - PER-T1-05 - Ellie GholamiEllie Gholami · DataEngBytes Perth 2023
• 27mDataEngBytes 2023 - SYD-T1-10 - Kanishka MohaiaKanishka Mohaia · DataEngBytes Sydney 2023
• 53mData Engineering and AIScale By The Bay 2018
• 26mDataEngBytes 2023 - SYD-T2-03 - Akanksha MalikAkanksha Malik · DataEngBytes Sydney 2023
• 02mDBTB INT Nemanja SpasojevicNemanja Spasojevic · Data By the Bay 2016
• 41mInfrastructure for ML ApplicationsNatalie Pistunovich · GoWayFest 4.0 (online edition)
• 21mDataEngBytes 2023 - SYD-T2-04 - Suneeta MallSuneeta Mall · DataEngBytes Sydney 2023