LLM Research Scientist (Pre-training & Computer Vision & Adversarial Robustness)
Mercor • Remote
Education
PhD
Type
Hourly
Pay
$100-120/hr
Your rate depends on your experience and the task.
Last checked
4d ago
Checked
Only apply if this role fits your skills and interests.
About this role
From the Mercor listing
We're looking for experienced machine learning researchers with hands-on experience training and improving deep learning models end-to-end, across vision and language. You'll work on well-scoped empirical open-ended ML research problems.
Responsibilities
- Train image classifiers and generative image models from scratch, and fine-tune open-weight language models.
- Get the most out of limited data, compute, and model-size budgets.
- Make models robust — to adversarial inputs and to adversarial conversations.
- Compress models to meet hard size and latency constraints without sacrificing accuracy.
- Diagnose and resolve training issues.
Requirements
We are looking for candidates with strong expertise in one or more of the following areas:
Adversarial Robustness
Experience with:
- Adversarial training of image classifiers (e.g. PGD-based training, TRADES).
- Evaluating robust accuracy under standard threat models (e.g. L∞ attacks, AutoAttack) and avoiding gradient-masking pitfalls.
- Managing the robustness–accuracy trade-off and robust overfitting.
Efficient Computer Vision
Experience with:
- Training image classifiers end-to-end, especially for fine-grained recognition (many visually similar classes, few examples per class).
- Model compression: quantization, pruning, and knowledge distillation from large teachers into small students.
- Deploying models under hard size or latency budgets (on-device, edge, or embedded settings).
Generative Image Modeling
Experience with:
- Training image generative models from scratch: diffusion models, GANs, VAEs, or flow-based models.
- Iterating against sample-quality metrics such as FID.
- Training-efficiency tricks that produce good generators quickly and at small parameter counts.
LLM Post-Training & Behavioral Robustness
Hands-on experience with one or more of:
- Supervised fine-tuning and preference optimisation (DPO, RLHF, RLAIF) of open-weight language models, including building your own datasets via synthetic generation, noisy or weak supervision, and rejection sampling.
- Shaping conversational behaviour over multiple turns: resistance to persuasion and sycophancy, calibrated confidence, and knowing when to accept corrections.
- Alignment-style fine-tuning that changes a specific behaviour while preserving general capability.
Multilingual Pre-training
Experience with:
- Training multilingual or low-resource-language models from scratch.
- Tokenizer design across scripts and typologically diverse languages.
- Balancing highly unequal per-language data (sampling temperatures, cross-lingual transfer) in data-constrained regimes.
Additional Areas of Interest
Experience in any of the following is a plus:
- Scaling laws and training-efficiency research.
- Curriculum learning and data ordering.
- Model evaluation: benchmark construction, contamination control, statistically sound comparisons.
- Uncertainty estimation and model calibration.
- Data augmentation and synthetic data for robustness.
General Qualifications
- 3+ years of machine learning research experience (PhD research counts toward this requirement).
- Strong experience with PyTorch, JAX, TensorFlow, or similar ML frameworks.
- Degree from a top-100 university, experience at a FAANG or comparable AI company, or an equivalent research track record through publications or impactful open-source contributions.
Why Join
- Work on cutting-edge machine learning research.
- Collaborate with leading AI researchers on challenging, high-impact projects.
- Flexible, project-based work with competitive compensation.
How long hiring takes
On the platforms we refer people to, half of hires came within about 25 days of applying. Time from applying to hire:
Talent pool members
When your profile is a strong match, we can put you forward to Mercor. Set up your profile first if you haven't.
Set up your profile →Research Scientist interview prep
Questions Research Scientist interviewers often ask. They aren't taken from this listing.
How do you design an experiment so that the result is interpretable whether or not the hypothesis holds?
I include positive and negative controls that tell me the setup worked, so a null result means the effect is absent rather than that the assay failed. I also decide the analysis in advance, including what counts as support, because an experiment that can only confirm is not a test of anything.
How do you pick which variables to control, randomize or record when a full factorial design is too expensive?
I rank factors by how plausibly they could confound the main effect, control the top few tightly, randomize the next group across runs, and record the rest as covariates. A fractional design then covers the interactions I care about, and the recorded variables let me check later whether something unexpected drove the result.
How do you decide which statistical test to use, and how do you avoid choosing one after seeing the data?
The test follows from the design and the outcome type, so I write it into the protocol before the first run, with the expected effect size and the alpha level. If the data later violate an assumption, I switch to the pre-specified fallback and say so in the paper rather than trying three tests and reporting the best.
What do you do to make sure another scientist could reproduce your analysis from your materials alone?
Every figure in my papers is generated by a script from raw data in a version-controlled repository, with a fixed random seed and a listing of the software versions used. Before submission I clone the repository to a clean machine and rerun it end to end, because that is the only way to find the dependency I forgot to document.
Why this role
This LLM Research Scientist (Pre-training & Computer Vision & Adversarial Robustness) role pays $100-120/hr for the STEM background you've already built.
Talent pool
We're light on STEM candidates
We've tracked 1,040 STEM listings and have few candidates to put forward for them. Set up a profile and we'll consider you for roles like this one.
Set up your profileSkills and categories
More jobs in these fields:
Related jobs
Verified reviews
Community reviews
Leave your review
Common questions
Does it cost money to apply to Mercor?
No, applying to and joining Mercor is free. Mercor charges the client a fee on top of your rate. If anyone asks you to pay to join, treat it as a scam.
Is Mercor for freelancers or full-time contractors?
Mercor places you with one client for a set contract, such as 'Python Tutor for 3 months'. Most roles are steady contract work rather than one-off tasks.
Is AI training work the same as traditional consulting?
It's closer to teaching. You grade the AI's logic, correct its mistakes and write the expert reasoning it lacks, instead of producing work for a client.
Why do expert AI training roles pay well?
The model already knows the basics. It needs help with hard edge cases, the judgment calls a senior professional in the field gets right.
What does STEM work look like for a LLM Research Scientist (Pre-training & Computer Vision & Adversarial Robustness)?
You check AI answers to STEM questions, correct their errors and write expert reasoning. As a LLM Research Scientist (Pre-training & Computer Vision & Adversarial Robustness), you're hired for that judgment, so hands-on STEM experience matters.
What specific skills does this listing call for?
Multilingual is named directly in the listing. If you don't have hands-on experience with this, expect the screening to test for it.
How much does this specific role pay?
This listing is posted at $100-120/hr, an hourly rate. Where you land in the range depends on your background and the assessment. Pay can change after we check the listing, so confirm the rate on the platform's application page.
What happens when I click Apply on this listing?
Apply opens Mercor's site, where you create an account and follow its sign-up steps.
Is a PhD required?
For this role, yes, or similar professional depth, and the assessment checks it. PhD candidates and people with published research have qualified without the degree.
How soon will I start working after applying to Mercor?
Usually after a few weeks. Applying puts you in Mercor's pool of candidates, and you start once a client, such as an AI lab, picks your profile.