#MLEjob #MLengineering #algorithms #careerchange #machinelearning #deeplearning
🔍 Are you feeling like your MLE job doesn’t have much ML? You’re not alone. Many MLEs work on algorithms that are more about engineering than machine learning. But don’t worry, there are steps you can take to transition to a role where you can spend more time working with traditional ML or deep learning.
Here’s a guide to help you navigate this transition and address your concerns about whether your experience will be considered “ML engineering” by other hiring managers.
## Understanding Your Current Role
First, let’s take a closer look at the aspects of your current job that are more focused on algorithms engineering rather than traditional machine learning. This will help you better articulate your skillset and experience to potential employers.
### Responsibilities in Your Current Role
In your current role as an “ML engineer”, you’ve been working on tasks that are more aligned with algorithms engineering, such as:
– Designing classifiers and estimators
– Handling large volumes of data stored in the cloud
– Defining data collection protocols
– Data visualization and feature extraction
– Creating Python processing pipelines for validation and tuning
– Implementing algorithms on-device in C++
– Validating algorithms with a combination of offline and on-device testing
### The Challenge
The majority of the algorithms you work on are hand-designed classifiers with manual rules and tunings, rather than traditional ML algorithms. Your role also involves working on a low-resource system with strict failure requirements, leading your team to avoid using ML unless absolutely necessary.
This has resulted in you only working on a few tasks related to ML algorithms in your time at this role, such as code generation, failure analysis, and creating a simple online linear model.
## Evaluating Your Concerns
You’ve expressed concern about whether your experience in your current role will be recognized as “ML engineering” by other hiring managers. Let’s address this and evaluate whether it’s something you should genuinely worry about.
### The Value of Your Experience
Remember, the saying “the algorithm is only 5% of the work” applies to many machine learning roles. Your experience in handling large volumes of data, defining protocols, and implementing algorithms on-device is incredibly valuable.
Furthermore, the approach your team takes to avoid using ML unless necessary can be an asset in roles that require a keen understanding of when to leverage machine learning and when to rely on simpler algorithms.
### Transitioning to a New Role
Moving into a role where you can spend more time with traditional ML or deep learning is entirely feasible. Your experience in algorithms engineering, coupled with your limited exposure to ML tasks, can position you as a candidate with a well-rounded skillset.
## Leveraging Your Experience for a Career Change
Now that we’ve addressed your concerns and highlighted the value of your experience, let’s explore how you can leverage your current role to make a successful transition to a more ML-focused position.
### Showcasing Your Skills
When applying for new roles, emphasize the following aspects of your current job:
– Your ability to handle large datasets and work with cloud-based data
– Experience in algorithm implementation and testing
– Adherence to strict failure requirements and avoidance of unnecessary use of ML
– Your involvement in tasks related to ML algorithms, such as code generation and failure analysis
### Upskilling Opportunities
Consider furthering your knowledge in traditional ML and deep learning through:
– Online courses in topics like neural networks, image recognition, and natural language processing
– Participating in ML-related projects or hackathons to gain hands-on experience
– Seeking mentorship from professionals in the field to expand your knowledge and understanding
### Networking and Job Search
Engage with individuals and communities within the ML space to expand your network and gain insights into potential job opportunities. Utilize platforms like LinkedIn to connect with professionals and explore open positions in companies that align with your career aspirations.
### Tailoring Your Resume and Cover Letter
Craft a compelling resume and cover letter that effectively communicates your transition goals and showcases your relevant skills and experience in algorithms engineering and limited ML tasks.
## Moving Forward with Confidence
It’s normal to feel uncertain about transitioning to a new role, especially when your current job doesn’t align perfectly with your desired career path. However, by recognizing the value of your experience and taking proactive steps to upskill and showcase your capabilities, you can confidently pursue opportunities in traditional ML and deep learning.
Remember, your journey in professional growth is a dynamic process, and your unique blend of skills and experiences will undoubtedly be valued in the right role. Don’t hesitate to embrace this transition and make the most of the opportunities that come your way.
In conclusion, don’t underestimate the worth of your algorithms engineering experience, and stay proactive in your pursuit of a more ML-focused role. With determination and a strategic approach, you can successfully make the transition and thrive in a career that aligns with your passion for machine learning.
You have a simple solution. As you know the theory and have the required education background and title.. all you need to do is land a new job. Given the position you are in and since most of the focus is on leetcode and system design and not actually what you worked on.. you can easily extrapolate and land a role that matches your objectives… You are thinking on right lines..
At the end of the day, the job is what it is. If you aren’t satisfied with it, the answer is always to find a job that needs you to do something closer to what you want to do.
This sounds like my dream job tbh. Any chance we can connect on LinkedIn? I wanna know a bit more if u dont mind sharing about your job.
This sounds a lot closer to ML “engineering” than using ML to recommend somebody songs on their phone.
A lot of the things you mentioned ARE machine learning. Unfortunately a lot of PhDs will tell you this is the reality: a lot of busy work and not so much designing novel algorithms. (side note: for this particular subfield of computer science, academia and industry are closely aligned and in sync, unlike other stuff like.. operating systems?)
You mentioned “on-device with significant compute/memory constraints” which is a pretty unique atmosphere to be doing machine learning compared to the more traditional “we’d like you to tune our models to increase viewer retention.” **That means it makes you stand out as a candidate**.
Either way, you might want to look for a more senior role at a smaller company or doing consulting work. That way you might have more say in what specific algorithms you’re using. Then you get to explore more options than just trees, deep learning and linear regression.
I don’t have much credibility to give you this advice but these are just my two cents.
One of the best parts of working at companies like FAANG is that you can switch teams or even job functions internally often without needing to interview. Maybe reach out to managers on other teams doing the kind of ML you are, tell them your situation and what you’re looking for, and see if they have room or would let you know when they do.
Doing it this way people won’t mind the lack of direct experience with what they’re working on (like the more ML heavy side of things) because they know you’ll make up with it knowledge about how the company’s internals and tools work and because it’s less hassle than hiring externally.
Basically how I switched from only doing Android my first two years at Amazon as a new grad to doing backend work
Honestly, your job sounds pretty typical. Model building is becoming just a few lines of codes so the world really needs people who can do the engineering and makes the models work. You can build an XGBoost model with like 5-10 lines of code. The hard part is handling the infrastructure which is what you do.
I think you might want to consider AI startups if you want to work on cutting edge ML/DL stuff.
You described most software jobs, that are just jobs
This is something I also do not understand at the current job market, there are many scenarios where the interview progress is different than the job itself.
One example is that in real life many of the ML algorithms works under vastly constrained environments, where CPU and GPU utilization is important, and you need to know TensorRT, C++, Python and such. However at the interview they dont really care about whether you did memory profiling to your C++ apps, everyone cares about the theory, but in reality you dont need the theory that much.
I don’t think that your job is very different than MLEs in general, but if you do want more ML/DL stuff, any kind of recommenders system teams might be worth considering for an internal transfer.
This is MLE and then some, both the things you say are unrelated to ML (they’re not) and those you say are. Algorithm research and building usually comes under a different title and, at least in my experience, is mostly done by PhDs.
My experience is at several startups and mostly in computer vision so ymmv.
If you don’t like your job… Then get a new one. What else do you want us to tell you?
OP, that’s the reality. That’s what ML usually looks like. I did ML research for my thesis, dealing with the algorithms is the smallest portion of the work. Mostly because I was solo and did everything myself. It’s funny how you find it boring, I enjoyed the engineering parts the most.
What you’re looking for is a data scientist position, especially on a team with large amounts of accessible data for DL. Usually they demand at least a masters in ML, more likely a PhD. That position deals more with the algorithms, they usually give the plumbing/engineering work to an MLE like you.
Why don’t you automate the whole process then and find a machine learning component from your present job?
Only way to show relevance of your function is to lead the initiative and use the company data to generate machine learning algorithms to solve business problems. Talk with department head. Ask them on 1-on-1 situation ‘if you could use data to solve a pressing problem now, what problem would you like solved?’ Then build a ML algo if possible or even preferred better solutions of tools better suited for the job.
Regardless, you have to maybe take initiative here and search out challenges and business problems