True ML Talks #22 |Deciphering ML and LLMs at Voiceflow

Описание: In the most recent edition of #TrueMLtalks, Denise the head of Machine Learning at Voiceflow. He works on how machine learning can be applicable to the company and building the infrastructure at Voiceflow. He Shared his valuable insights on the process of establishing machine learning teams and developing natural language understanding (NLU) driven solutions, as well as conversational AI solutions from scratch.

We spoke on so many intriguing topics which included the following:
1.Denys's passion for machine learning
2.Denys’s journey in establishing Voiceflow's ML infrastructure and team
3.In-depth discussion on the impact of recent GPT models on the ML landscape.
4.Voiceflow's varied industry clientele and their approach to addressing associated challenges.
5. Challenges when Voiceflow shared deployment templates with data scientists."
6. Use cases demonstrating backfilling's benefits in downtime
7.The technological solutions employed by Voiceflow to implement autoscaling
8.The cloud service provider utilized by Voiceflow
9.Denise's insights into the adoption of Karpenter or Autopilot within Voiceflow
10.Voiceflow’s choice of specific model server after testing various latency-sensitive models.
11.Voiceflow’s experimentation with open AI models Like Llama or Falcon
12.Voiceflow's approach to fine tuning the GPT model
13.What difficulties arose when utilizing GPT-4 due to its significantly greater scale?
14.The comprehensive details regarding the development of RAG systems at Voiceflow
15.The obstacles encountered by Voiceflow when shifting from NLP-based solutions to alternative models."
16.Voiceflow incorporation of OpenAI costs into their pricing strategies.
17.The technologies embraced by Voiceflow to ensure data privacy and protection
18.The process for accommodating a custom model within Voiceflow
19.Voiceflow's favored data format: Text or Visual

00:15: Start
00:16: Introduction
1:21: Denys's machine learning passion and career journey
02:45- Denys's journey building Voiceflow's ML infrastructure and team
06:43: In-depth talk on recent GPT model impacts on ML
08:55 : Voiceflow's diverse clients and their approach to industry challenges.
13:04: Challenges of Voiceflow sharing deployment templates with data scientists
17:43 : Voiceflow's tech for implementing autoscaling.
20:48 The cloud service provider utilized by Voiceflow
21:06: Denise's take on using Karpenter or Autopilot at Voiceflow
22:13: The model server that worked for Voiceflow after latency-sensitive model tests
23:56 Voiceflow's OpenAI model experimentation
25:33 : Voiceflow's approach to fine tuning the GPT model
27:24: Challenges with GPT-4 due to its larger scale
28:53: Details on developing RAG systems at Voiceflow.
30:53: Challenges when transitioning from NLP to other models at Voiceflow
34:00 Voiceflow incorporation of OpenAI costs into their pricing strategies.
35:13: Technologies for Voiceflow's data privacy and protection
37:42 : The process for using a custom model on Voiceflow
38:37: All about data types at Voiceflow

ABOUT OUR GUEST

Denys assumes the role of leading the machine learning team at Voiceflow, where he has been an integral part of the team for nearly two and a half years. He initially joined as the founding ML engineer, and during her tenure, he has been instrumental in identifying the various ML applications that can be employed within the company. Denise played a pivotal role in devising the strategy, assembling the team, and overseeing the growth of several product lines within the machine learning domain.

Before his time at Voiceflow, Denise served as a senior cloud architect for a global bank, focusing on data systems, MLOps, and core infrastructure development.

Visit the following URL to get in touch with Denys: -   / denyslinkov  

ABOUT OUR CHANNEL
In our video series TrueMLTalks, we speak with experts in the machine learning field from organisations including Gong, Intuit, SalesForce, Facebook, DoorDash, and others. It is a useful tool for professionals wishing to remain up to date on the most recent developments in the field because we give an informed view of their experiences managing complex ML pipelines and creating successful best practices.

ABOUT TRUEFOUNDRY
TrueFoundry is a PaaS for cross-cloud machine learning deployment that enables businesses to speed up model testing and deployment while preserving total security and control for the Infrastructure/Development Operations team. We give machine learning teams the ability to deploy and monitor models with 100% reliability and scalability in just 15 minutes, saving money and allowing models to be put into production quicker, which generates real business value. In order to protect data privacy and other security issues, we deploy on the customer's infrastructure.

A quick platform demo of mins : https://t.ly/0fk1
Browse our blog: https://blog.truefoundry.com/
Subscribe to our newsletter: https://t.ly/oORp