Data sci at WhyLabs, UW eScience Institute, UW iSchool PhD student. Robust eval, responsible ML/AI, synthetic data, & model monitoring over time.
Seattle, WA
- Tweets 668
- Following 596
- Followers 819
- Likes 864
Want to compare data-centric and model-centric approaches for production AI? On Monday at 9am, @zalessya and I from @WhyLabs will be presenting an @OReillyMedia course on just that topic. We've also added a look at LLMs. oreilly.com/live-events/appl…
3
250
So cool to stumble upon highlights from my little brother and his teammates. Since when did my old high school get a media day!? Fancy.
Highlights from the Reeths-Puffer boys varsity basketball team’s media day yesterday at CatchMark. Thanks to Farm Bureau Insurance - McCleve Agency for the sponsorship!
2
309
Join Alessya's LLM security workshop in 30 minutes! She's an excellent speaker and teacher with lots of tricks and best practices up her sleeves. I'll be listening in as well.
We’re going live at 10 am PST for an Intro to #LLMSecurity Workshop! Don’t miss the chance to learn how to implement monitoring for common security issues and adopt best practices to ensure your #LLMs are safe and responsible!
bit.ly/3MMgPN3
#ML #OWASP #ResponsibleAI
343
Bernease Herman retweeted
New short course! Quality and Safety for LLM Applications, created with @WhyLabs (an AI Fund portfolio company) and taught by @bernease, shows how you can mitigate hallucinations, data leakage, and jailbreaks. Come learn more in the course, available now! deeplearning.ai/short-course…
14
130
601
105,939
Bernease Herman retweeted
O’Reilly's AI Superstream is happening NOW! Discover the untapped potential of data-centric AI with expert-led sessions including @bernease’s talk on #LLM Observability at 11:05 am PDT!
Sign up for a free trial to access the event: bit.ly/467mmoK
1
4
332
AI Superstream on Data-Centric AI is getting started right now. I'll be joining at 11am PT to talk about the relationship between LLMs & DCAI that we utilize at @whylabs, but great speakers on other topics now.
Sign in, or sign up for your free trial: oreil.ly/NLdJ-E
1
7
1,187
Bernease Herman retweeted
Gaming, product design, marketing assets, mobile apps...come share what you’re building with other #AI developers in the Seattle area!
Join us for an after-hours meetup next week with @OctoML, MLC & others!
Register now to save your seat: lu.ma/mt0c5sdl
#ML #MLOps
1
4
596
Join me tomorrow to talk about monitoring massive and unstructured ML datasets in production!
Attending @PyDataSeattle? Don’t miss these sessions:
📍Monitoring in the era of #GenerativeAI, LLVMs, and embeddings | April 27 @ 3:30 pm | bit.ly/3NhaeLI
📍Being well informed: Building a #ML Model Observability pipeline | April 26 @ 1:30 pm |
bit.ly/3oFo549
ALT PyData Seattle
1
4
325
Study reveals the optimal time to work out to live longer nypost.com/2023/02/23/study-… via @nypost
1
146
At AI at Scale Summit hosted by @AiInfra! @WhyLabs session on Truly Scalable Monitoring for NLP and CV Datasets is happening in 20 minutes! Not too late to sign up: ai-infrastructure.org/ai-at-…
Now: Networking session #2
10am PT / 1pm ET: Our session, in Track 2
1
4
243
Bernease Herman retweeted
Hear how @Shopify is streamlining their #ML workflow, cutting down on engineering time, and ensuring model performance!
Check out @iambargar's session on the role of #FeatureStores in data quality assurance during the Robust & Responsible AI Summit: bit.ly/3I9GCgo
3
3
532
Andre gives clear and helpful presentations. Be sure to check out his webinar on Wednesday!
Don’t miss the live webinar with @useautomation on detecting data quality and drift in your production #ML stack! Register now 👇
Wed, Feb 1 | 10:00 am PST
eventbrite.com/e/whylabs-dem…
#MachineLearning #MLOps
206
Would love to grab coffee with folks heading to New Orleans for #neurips2022 next week. I'll be there Tues Nov 29 - Fri Dec 2. Send me a message and let's arrange a time to chat about #mlops and production ML. Excited to be in person once again!
4
Bernease Herman retweeted
Join @zalessya's tech talk at #DevFestSEA22 on #MLObservability, including different approaches to instrumenting models and pipelines, collecting actionable telemetry, and catching model failures!
Get your free ticket here: bit.ly/3TIdTln
👉 Saturday, 11/19 @ 10AM
ALT DevFest 2022 in Seattle
3
6
I bought a pair of @VessiFootwear a few months back in order to test their suitability in Seattle weather.
I can't tell you about the waterproof ability of the shoes, but somehow every time I wear them, the rain clears up. So a win for me.
1
2
Bernease Herman retweeted
The #DataCentricAISummit from @AiInfra kicks off on Sept 29 with an incredible lineup of some of the best data-focused practitioners in the world 🤩 including @WhyLabs Sr. Data Scientist @Bernease Herman!
Secure your seat now. It’s FREE and all online: bit.ly/3ATC4ak
ALT Data-Centric AI Summit
1
6
Is the uncanny valley effect (and thus, the required generative image quality) stronger for video than still images? I assumed you'd need extremely high quality video to not creep people out to the point of it being commercially non-viable now.
1
2
Bernease Herman retweeted
1921
ALT There is no question that the youngsters of 1921 are setting too fast a pace. It is bad for their health and dangerous to their morals when children who can hardly "toddle" are doing it six nights out of seven. There are too many high powered cars parked around the high schools. The kids have it "too soft" for their own good. If those Kansas City mothers will just come to an agreement to see it that their boys and girls are soundly spanked until they are made to like it, they will have little further cause for worry. As matters stand the kids have them bluffed and they are foolish enough to admit it. --The Laclede County Republican, 4 Feb 1921
31
141
3,672
Bernease Herman retweeted
I propose that we adopt the term "Large Self-Supervised Models (LSSMs)" as a replacement for "Foundation Models" and "LLMs". "LLMs" don't capture non-linguistic data and "Foundation Models" is too grandiose. Thoughts? @percyliang
38
42
517
Bernease Herman retweeted
Is your team flying blind? Many teams test their #MachineLearning models offline but conduct little to no evaluation after deployment.
Proactively detect and respond to model performance changes before your customers or stakeholders even notice an issue: bit.ly/3zT77CM
5
10