Report this Job
ML Ops Engineer
Columbus, Ohio
6+ Months
Description:
- AWS cloud certification
- Docker, Sagemaker, Container
- Hadoop Cluster for Data lake and production runs of models
- This is done via batch not real-time
- Weekly, daily, and monthly updates to the data model runs and usage daily
- Migrating away from Hadoop to AWS see it in 2 ways data lake aspect (well handled by the current team) data science piece (MLOps is where they are needing help)
- MLOps is on the horizon, but more development needs to be done Containers, Terraform, CloudWatch
- Needing someone with a strong drive to learn in the MLOps space open to a mid-level candidate with the hunger to learn more
- Ultimately a strong AWS engineer with a focus on the below areas
Must Haves:
- CloudWatch
- Sagemaker
- Python
- Containers Docker and Kubernetes
Screener:
Responsible for / Working On:
- Had a big cluster of CPUs stored a bunch of data, served as a place where data scientists would build their models and put them into PROD
- Wasnt scalable
- May be getting data from some unique place, have a favorite software, or even favorite sub-packages within certain software, and all this uniqueness leads to a lot of inefficiencies because nobody is doing the same thing even if Bill builds something great with code, it can't be used by Susan because any one of the above reasons.
- Need to take away approach where they are building model 1, deploy it, build model 2 deploy it, etc.
- This plays into what is called MLOps
Instagram, an example of how various parts of the app are actually different containers behind the scene that are then clustered out based on the volume of people trying to access different features of the app at different times.
To run models may require a large number of containers that essentially feed and deploy that back to the end users.
Meta, LinkedIn, laand rge tech companies are well ahead of banks and finTechs.
Hadoop data model and storage
need to focus n making the models. About how to make all the models efficiently in order to run all at the same time.
Key need currently is focused on AWS, and having key understanding of containers (Docker-this is key), Sagemaker (AWS analytics modules).
MLOps need to come from this background.
Marketing example, this person has $123k will send them a targeted message to open a savings account and the bank will give them $500
Model monitoring is also crucial to ensure what is running is actually effective, if not, what do we need to change then redeploy.
LinkedIn, different models would be things like DMs, jobs, and various different apps within an app.
Saksoft