How do you build machine learning systems in healthcare without a medical degree?
In this talk, Vladimir Kirilenko, Senior ML Engineer at deepsense.ai, shares his engineer’s-eye view of medical AI:
- where the data really comes from,
- why metadata can’t be trusted,
- what happens when abbreviations, negations, and speech-to-text collide,
- why teeth shine under X-rays, and
- how to survive artifacts, bloomy scans, and gigabytes of CT data.
This isn’t a dive into new models or SOTA papers – it’s a survival guide for ML engineers who might find themselves in healthcare projects.
Timeline
00:00 Intro & agenda
1:31 What is Medical AI?
2:42 Pharma ML vs. Clinical ML
4:15 How medical data is created
7:01 DICOM standard: the good, the bad & the messy
9:29 Metadata pitfalls: creative mappings & dataset traps
15:00 Clinical text challenges
22:06 Computer vision in medicine
29:33 Recap & survival tips for engineers
Speaker
Vladimir Kirilenko
Senior ML Engineer at deepsense.ai
