Tickets

Loading…

Zainab Bawa

Zainab Bawa

@zainabbawa

Kannan S

@skannan Editorial Assistant

Nischal HP

Nischal HP

@nischalhp Editor

MLOps 2022 conference report

Submitted Nov 30, 2022

Context

This is a report of the MLOps 2022 conference, held on 11 November, at the Microsoft Reactor, Bangalore.

250 persons registered for the conference. These included practitioners from ML, data science and data engineering teams. Enterprises, unicorn startups, and early stage startups and SaaS companies equally participated in the conference.

100 participants showed up at the conference, in person. Some of the participants who could not show up at the venue purchased a subscription to The Fifth Elephant to access the videos after the conference.

Participants attended with the conference to learn from the experiences of organizations who have implemented MLOps workflows. Participants gained much value from speakers’ sharing real-life case studies and organizational experiences.

MLOps 2022 conference was the first in-person event hosted by The Fifth Elephant after the pandemic. In 2021, The Fifth Elephant hosted MLOps 2021 conference

Participant testimonials

  1. Aparajitha Murali’s takeaways from MLOps 2021:

Here are my takeaways from The Fifth Elephant #MLOps 2022 day talks by Hasgeek on 11.11.22 :

✅How to have more accountability and better governance in #ML projects
✅Focusing on realistic timelines to deliver ML projects that have a higher #business impact and build as per business objectives.
✅Looking into SubML (an approach for lightweight, faster and evolving use cases of ML) because not all of us need or have access to large #datasets and the resources to work with them. But there are plenty of #SubML problems to solve
✅How the formation and existence of silos can be detrimental to a business and the people involved
✅About inference #graphs, building #featurestores, working with #YAML files, data versioning (like versioning your code!), data meshes, and transparency.
✅How important it is to #log every darn thing. [I agree because it will save your neck down the line. Multiple times.😅]
✅How MLOps could be built using the existing fundamental principles of #DevOps
✅The importance of asking questions about Personal Identifiable Information (#PII): Do we really need to collect PII to accomplish our objectives? Or are there other solutions? How to scan for sensitive information in data before it flows into your system etc.?
✅Focusing on and building #explainability into data upfront.
✅AI fairness and responsible #AI

  1. From participant Manu Awasthi:

Super interesting to see the kind of complex ML pipelines that folks are building, and the mind boggling set of tool choices that are available to create them.

  1. Observations posted by participant K V Sandeep Moudgalya, commenting on each talk.
  1. “Developing ML Applications” by Venkata Pingali, Malavika Lakireddy and Saransh Verma
    Excellent interaction by the trio on how to bring in best MLOps practices to make real business sense for AI/ML application consumers. Stressing on how TerraPay and Scribble Data has been successful in developing business focused ML applications in collaboration.
  2. “How PayPal uses Inference Graphs to design ML systems” by Sharmili Srinivasan and Saurav Raj
    Very insightful interaction on how PayPal is employing Inference Graph appoach to build ML systems using Seldon as a major MLOps platform.
  3. “War stories on MLOps and data governance - learnings from Data Mesh” by Madhusudhana Rao Podila
    Insightful talk on how transparency and visibility can brought into ML practices by employing Data Governance approaches using DATA mesh
  4. Panel discussion on “Building value to MLOps with data governance” by Ashwin Kumar, Balvinder Kaur Khurana, Sathish K S (Zeotap) and Srinivasa Rao Aravilli
    Thought provoking interaction on bringing in data governance in MLOps practices and the challenges that come along with that in its implementation.
  5. “Make AI real - from labs to production” by Meduri Ravi Kumar
    A fresh perspective on the challenges faced by AI/ML projects in getting it going and how MLOps practices can help in facing those challenges.
  6. “ML at scale: how Udaan built its ML platform” by Sai Sharan Tangeda and Sampan S Nayak
    Excellent talk by the duo on how Udaan employed Kedro and BentoML based ML frameworks to systematize their MLOps practices.
  1. Speaker Sai Sharan Tangeda’s testimonial, thanking participants for their generous compliments towards Udaan’s talk at The Fifth Elephant on ML Platform and how it has enabled Data Teams at Udaan.

Photos from the conference are published here.

Editorial and speakers

MLOps 2022 was curated by Nischal HP and S Kannan. The Fifth Elephant team reached out to speakers from the 2021 edition to follow up on their talks and journeys with MLOps.
Outreach was also done with an extended network of speakers, aggregated via Hasgeek’s brands, namely Rootconf, Privacy Mode and Anthill Inside.
The editors reviewed speakers’ slides and did one rehearsal of their talk before the conference.

The following speakers participated in the conference (in the order of the talks):

  1. Dr. Venkata Pingali - CEO and co-founder at Scribble Data
  2. Saransh Verma - Director of Analytics at TerraPay
  3. Malavika Lakireddy - Head of Product at Zeotap
  4. Sharmili Srinivasan - ML engineer at PayPal
  5. Saurav Raj - ML engineer at PayPal
  6. Madhu Podila - Data strategist at Thoughtworks
  7. Ashwin Kumar - Head of Machine Learning, Borneo
  8. Balvinder Khurana - Principal Consultant at ThoughtWorks
  9. Sathish KS - CTO at Zeotap
  10. Srinivasa Rao Aravilli - Director of Data and AI Platform at Visa
  11. Ravi Kumar Meduri - Executive Vice President at Innominds
  12. Sai Sharan Tangeda - Product engineer at Udaan
  13. Sampan Nayak - Full stack engineer at Udaan

Sponsors

The following organizations participated as sponsors for the conference:

  1. Scribble Data - primary sponsor
  2. Thoughtworks India - community sponsor
  3. Microsoft Reactor - venue sponsor

Outcomes from the conference

One of the key outcomes of the MLOps 2022 conference was the launch of Hasgeek subscriptions. MLOps 2022 saw an uptake in subscriptions, and feedback from participants on what they expect from subscriptions.

The other outcome from MLOps 2022 was lead generation for potential partnerships with organizations such as Innominds, PayPal and Ashoka University.

The Fifth Elephant has also collaborated with Thoughtworks to evangelize Data Mesh framework as an alternative to data management and data governance in organizations.

Future plans

MLOps will launch tracks covering the following themes:

  1. Economies of data science
  2. Running Machine Learning in production.
    Activities such as panel discussions and talks will be held to provide coverage to the themes.

To stay updated, subscribe to hasgeek.com/fifthelephant

Comments

{{ gettext('Login to leave a comment') }}

{{ gettext('Post a comment…') }}
{{ gettext('New comment') }}
{{ formTitle }}

{{ errorMsg }}

{{ gettext('No comments posted yet') }}

Hybrid access (members only)

Hosted by

All about data science and machine learning

Supported by

Sponsor

Scribble Data builds feature stores for data science teams that are serious about putting models (ML, or even sub-ML) into production. The ability to systematically transform data is the single biggest determinant of how well these models do. Scribble Data streamlines the feature engineering proces… more

Promoted

Community Sponsor

Thoughtworks is a global technology consultancy that integrates strategy, design and engineering to drive digital innovation.

Venue Sponsor

LEARN, CONNECT, BUILD.