← Back to Blog

Machine Learning Engineer Resume for New Grads: What to Include and What to Skip

Published ยท Finch Team

Updated

The demand for ML engineers is real. If you're a CS grad who's taken machine learning courses, built a few models, and done something with PyTorch or scikit-learn, you're in a decent position, but only if your resume actually communicates that work clearly.

The problem most new grad ML resumes have: they list a bunch of frameworks but don't show what was actually built or what the results were. A hiring manager reads "PyTorch, TensorFlow, scikit-learn, Pandas" and has no idea if you can build anything useful or if you just watched some YouTube tutorials.

Here's how to build an ML-focused resume that actually gets past that.

The Fundamental Difference Between an ML and a General SWE Resume

A general software engineering resume emphasizes what systems you built: the scale, the impact, the technologies used. An ML resume needs to show all of that and also answer a different question: did the model actually work?

Specificity is everything. "Built a sentiment analysis model" is vague. "Built a sentiment analysis model using BERT fine-tuned on 50k labeled tweets, achieving 89% accuracy on a held-out test set, deployed via FastAPI." That's a resume bullet. The number matters. The deployment step matters. The methodology matters.

ML is a field where a lot of people have done toy projects with off-the-shelf datasets and never shipped anything. If you've done something more than that, you need to make that visible immediately.

What to Put in Your Technical Skills Section

Your technical skills section for an ML role should be organized differently than a general SWE resume. Group it by function, not alphabetically.

A strong ML skills section looks like:

Languages: Python, SQL, C++
ML Frameworks: PyTorch, TensorFlow, scikit-learn, Hugging Face Transformers
MLOps/Infrastructure: AWS SageMaker, Docker, MLflow, FastAPI, S3
Data: Pandas, NumPy, Spark (basic), dbt

Things to include that most new grad ML resumes miss:

  • MLOps tooling, even at a basic level (MLflow for experiment tracking, Docker for containerization, any deployment experience)
  • Cloud platform experience (AWS, GCP, or Azure: even free tier projects count if you deployed something)
  • Data pipeline tools: even basic Pandas and SQL show you understand the full data-to-model workflow

Things to leave out:

  • Tools you only used once in a tutorial and would need to relearn from scratch
  • Long lists of libraries that any Python developer would know (requests, json, os)
  • General CS skills that aren't ML-relevant on an ML-focused resume (listing "HTML, CSS" when applying to ML roles is noise)

How to Write Strong ML Project Bullets

This is where most new grad ML resumes fall short. Here's the template that works:

[What you built] + [method/model used] + [dataset/scale] + [result/metric] + [what happened to it after training]

That last piece (what happened after training) separates people who have real experience from people who only ran notebooks. Did you deploy it? Write an API around it? Use it to power anything? Even a basic FastAPI endpoint that serves predictions locally counts. Add it.

Examples:

Weak: "Implemented a neural network for image classification using PyTorch"

Strong: "Fine-tuned ResNet-50 on a custom dataset of 12,000 labeled product images, reaching 94% top-1 accuracy, served predictions via a FastAPI endpoint containerized with Docker"

Weak: "Built a recommendation system for a class project"

Strong: "Built a collaborative filtering recommendation system on 100K MovieLens ratings using matrix factorization (SVD), achieving RMSE of 0.87, deployed as a Flask API with a React demo frontend"

The second versions answer every question a recruiter or ML engineer would ask: what kind of model, what data, how did it do, did you actually use it for something.

The Projects That Matter Most Right Now

If you're building new projects to strengthen your ML resume in 2026, here's where to focus:

LLM integration work. Building something that uses an LLM API (OpenAI, Anthropic, or open-source models) plus retrieval or tool use is now a baseline expectation at many companies. This doesn't have to be complicated: a RAG system over a document corpus, a simple agent with a few tools, or a fine-tuning experiment on a domain-specific dataset all demonstrate current skills.

MLOps basics. Training a model is table stakes. Showing you can track experiments (MLflow or Weights & Biases), version datasets, or containerize a model for deployment demonstrates you understand the production side of ML, which is where the actual job happens.

A deployed project. Something people can actually interact with is usually more compelling than a notebook alone. It doesn't have to be at scale. A free-tier Hugging Face Space, a Streamlit app, or a simple API endpoint that you can demo in an interview is evidence you can ship, not just experiment.

How to Handle an ML-Heavy Coursework Section

If you have strong ML coursework (a graduate-level ML or deep learning course, a computer vision or NLP course), it's worth calling that out more specifically than just listing your degree.

You don't need a separate "Relevant Coursework" section. Instead, either:

  1. Mention specific courses in your education section: "Relevant: Deep Learning, Natural Language Processing, Probabilistic Graphical Models"
  2. Pull the best project from a course into your projects section with a note that it was coursework

The second approach is better because it shows the actual work, not just the class name. A course project where you got 91% on a benchmark is more convincing than listing "Machine Learning" as coursework.

What ML Recruiters Are Actually Screening For

A few things that come up consistently in ML interview processes:

Python fluency is assumed. If you're not fast and comfortable with Python, ML roles are going to be hard. Interviewers expect you to write readable, working Python quickly without looking things up.

Understanding of the basics. You'll get asked about gradient descent, backprop, bias-variance trade-off, and regularization. Not to prove you have the textbook memorized, to check that you understand what's happening inside the models you're using. Know these cold.

Practical ML judgment. Questions like "your model is overfitting, what do you do?" or "how would you handle class imbalance in this dataset?" These are practical, open-ended, and they're how interviewers separate people who've run through tutorials from people who've actually debugged ML systems.

A specific system you've built. Many ML interviews ask you to walk through something you've built in detail. Know your best project deeply: the data pipeline, the model choices, the results, what broke, what you'd do differently.

ML Resume vs. General SWE Resume: Which One to Send?

If you're applying to both ML and general SWE roles, you should have two different resumes: not two completely different documents, but two targeted versions.

Your ML resume leads with ML projects, puts ML frameworks first in your skills section, and frames your work in terms of models and metrics. Your SWE resume leads with system-building work, emphasizes scalability and engineering craft, and keeps ML as a supporting skill rather than the headline.

Sending the same resume to both cuts your match rate on both. The keyword sets are different, the framing is different, and what the recruiter is looking for is different.

Keep two base versions of your resume, one for each track, so you can apply to an ML-focused role at one company and a general SWE role at another in the same session without rewriting everything. Finch's board lists open roles for both, newest first.

ML Resume Checklist

Projects section:

  • Every ML project bullet answers: model type, dataset size, metric, and what you did with it after training
  • At least one project has been deployed or has an accessible demo
  • At least one project demonstrates LLM or modern NLP work if you have it
  • Numbers appear in every project bullet (accuracy, dataset size, RMSE, inference time)

Skills section:

  • Organized by function: languages, ML frameworks, MLOps/infra, data tools
  • MLOps tooling included even if only at a basic level
  • Cloud/deployment experience called out
  • No padding: only tools you can discuss in an interview

Coursework:

  • Strong ML courses called out explicitly in the education section or as project entries
  • Course projects where you got strong results moved into the projects section

Targeting:

  • ML resume version leads with ML-specific work and frameworks
  • Different version ready for general SWE applications

Entry-level ML hiring is competitive, but the bar for having a strong project-based resume is achievable. One well-documented deployed project is worth more than a long list of frameworks. Focus on building one thing all the way through (training, evaluation, deployment) and make sure your resume tells that full story.