AIWorkJobs
Mercor logoMercorPlatform listing

LLM Research Scientist (Pre-training & Computer Vision & Adversarial Robustness)

Tackle open-ended empirical research across language and vision, from training image models from scratch to fine-tuning open-weight LLMs. The work calls for hands-on research experience and practical judgment under…

Remote · Eligibility not specifiedHourly
Work arrangement
Remote
Eligibility
Eligibility not specified
Engagement
Hourly
Compensation
$100–$120/hr
Weekly commitment
40 hours per week

About the role

Tackle open-ended empirical research across language and vision, from training image models from scratch to fine-tuning open-weight LLMs. The work calls for hands-on research experience and practical judgment under tight data, compute, model-size, and latency budgets. You should bring depth in at least one of the listed research areas, such as adversarial robustness, generative modeling, multilingual pre-training, or LLM post-training.

What you would do

  • Train image classifiers and generative image models from scratch, and fine-tune open-weight language models.
  • Design experiments that make effective use of limited data, compute, and model-size budgets.
  • Improve model resistance to adversarial inputs and manipulative or adversarial conversations.
  • Compress models to meet strict size and latency limits while preserving accuracy.
  • Identify and fix problems that arise during model training.

Requirements

  • At least 3 years of machine learning research experience; PhD research counts toward this threshold.
  • Hands-on experience training and improving deep learning models from end to end across vision or language.
  • Strong knowledge of one or more areas covered by the role, such as adversarial robustness, efficient computer vision, generative image modeling, LLM post-training, or multilingual pre-training.
  • For adversarial robustness work: experience with image-classifier adversarial training, such as PGD or TRADES, and robust-accuracy evaluation with threat models such as L∞ attacks or AutoAttack. You should also understand gradient masking, robust overfitting, and the robustness-accuracy trade-off.
  • For efficient computer vision work: experience training image classifiers end to end, especially for fine-grained recognition with similar classes and few examples per class; model compression through quantization, pruning, or knowledge distillation; and deployment under strict size or latency limits.
  • For generative image modeling work: experience training diffusion models, GANs, VAEs, or flow-based models from scratch, improving sample quality against metrics such as FID, and producing capable generators with limited training time or parameter counts.
  • For LLM post-training work: hands-on experience with supervised fine-tuning or preference optimization methods such as DPO, RLHF, or RLAIF for open-weight models. Experience may include creating datasets through synthetic generation, noisy or weak supervision, or rejection sampling.
  • For LLM behavioral robustness work: experience shaping multi-turn behavior, including resistance to persuasion and sycophancy, calibrated confidence, and appropriate responses to corrections; alignment-style fine-tuning that changes a specific behavior while preserving general capability also qualifies.
  • For multilingual pre-training work: experience training multilingual or low-resource language models from scratch, designing tokenizers across scripts and typologically diverse languages, and balancing unequal language data through methods such as sampling temperatures or cross-lingual transfer.
  • Strong experience with PyTorch, JAX, TensorFlow, or a comparable machine learning framework.
  • A degree from a top-100 university, experience at a FAANG or comparable AI company, or an equivalent research record shown through publications or impactful open-source work.

Preferred qualifications

The source marks these as preferred, not required.

  • Research in scaling laws or training efficiency
  • Experience with curriculum learning or data ordering
  • Model evaluation, including benchmark design, contamination control, or statistically sound comparisons
  • Uncertainty estimation or model calibration
  • Data augmentation or synthetic data for robustness

Who should apply

Machine learning researchers with at least 3 years of research experience, strong framework skills, and hands-on depth in one or more listed areas should consider applying. Candidates can meet the research-track qualification through a top-100 university degree, work at a FAANG or comparable AI company, or an equivalent record of publications or impactful open-source contributions.

How applying works on Mercor

Create a profile, complete the role checklist and AI interview, then verify your identity before a project offer.

  1. Create an account and confirm your résumé.
  2. Complete the role checklist, written assessment, and AI interview where the listing requires them.
  3. Submit the application and wait for review or matching.
  4. Complete a work trial and identity verification before onboarding.

Typical timeline: The AI interview is about 20 minutes; work trials are often 4–6 hours; identity checks are usually minutes.

Paid steps: The work trial is paid.

Assessment and interview requirements vary by role. Check the checklist inside the specific listing.

Process checked Nov 19, 2026. AIWorkJobs does not submit applications or charge a fee. Read the full Mercor process and its official source on the Mercor page.

Skills mentioned by the source

  • machine learning
  • deep learning
  • pytorch
  • jax
  • tensorflow
  • adversarial training
  • computer vision
  • image classification
  • model compression
  • quantization
  • pruning
  • knowledge distillation
  • diffusion models
  • generative adversarial networks
  • variational autoencoders
  • flow-based models

Listing reference: llm-research-scientist-pre-training-computer-vision-adversarial-robustness-ZGvpVY. Source: Mercor. If something here contradicts the source, the source controls. Report a correction.

Related roles

Other work on the index.

See more in this category

micro1

Member of Technical Staff, Model Training (post-training research)

Explore how algorithms, training data, and evaluation shape reasoning, generalization, and agent capabilities in large language and multimodal models. You’ll test post-training approaches such as reinforcement learning, distillation, and data-efficient training, then turn research findings into scalable methods. The role spans experimental machine learning research and large-scale model development, from forming hypotheses to validating results.

Location and eligibility
Remote · Eligibility not specified
Engagement type
Full-time
Compensation
$7–$8/hr
  • Reinforcement Learning & Optimization
  • Post-Training Algorithms
  • Training Data Strategy & Curriculum Learning
AI trainingView job

Turing

LLM Java Developer

Develop and maintain Java code for training and optimizing AI models.

Location and eligibility
Remote · Eligibility not specified
Engagement type
Contract
Compensation
Pay not listed
  • java
  • large language models
  • llm training
Machine Learning & AIView job

Mercor

MLOps Engineer, LLM Performance (Kernels, Profiling, Serving)

Build and assess technical training material for frontier AI systems, with a focus on LLM infrastructure. You’ll create and review tasks involving accelerator kernels, performance profiling, distributed workload debugging, and high-throughput inference. The work also supports model training and evaluation. The role is fully remote and open worldwide, with a 40-hour weekday commitment and no other engagements during the work period.

Location and eligibility
Remote · Eligibility not specified
Engagement type
Hourly
Compensation
$90–$120/hr
  • mlops
  • ml systems
  • pallas
AI trainingView job

Handshake AI

Visual Design Professional

As a Handshake AI Fellowship participant, you’ll help assess and develop how large language models handle visual storytelling, brand design, and professional communication. Work remotely on flexible, self-scheduled project assignments with no minimum hour commitment. Projects may involve writing design-focused prompts, reviewing model responses, and researching topics with AI. Placement depends on project availability, and the fellowship runs year-round with openings that vary by design specialty.

Location and eligibility
Remote · United States only
Engagement type
Part-time
Compensation
$100–$125/hr
  • visual design
  • brand design
  • product design
Design & CreativeView job