NLP Data Scientist

Zoho; United States; Remote; USD 120,000 - 190,000; Part time

Job summary

Fyerx seeks an experienced NLP Data Scientist / LLM Fine-Tuning Specialist to own open-source model optimization tracks. You will drive PEFT-based fine-tuning, curate high-quality datasets, and implement RLHF/DPO to enforce safe, domain-aligned outputs.

Responsibilities include scaling multi-GPU training, model quantization, and collaborating with MLOps for cloud deployment and inference serving. Remote offshore contract role with international scope.

Qualifications

• 4 to 8 years of core data science or advanced ML engineering experience with 3+ years actively training, evaluating, and fine-tuning NLP systems.
• Expert-level mastery of Python, PyTorch, transformers (Hugging Face), and vector calculations.
• Deep understanding of attention mechanisms, tokenization constraints, context window effects, loss optimization, and CUDA hardware.

Responsibilities

• Lead LLM fine-tuning initiatives using PEFT techniques (LoRA, QLoRA, Prefix/Prompt Tuning) on open-source architectures (Llama, Mistral).
• Curate, clean, structure high-quality training datasets with deduping, tokenization schemes, synthetic data pipelines, and human-in-the-loop validation.
• Implement RLHF or DPO to align models with safety, helpfulness, and tone guards.
• Optimize model footprints with post-training quantization (GGUF, AWQ, GPTQ) to reduce compute budgets.
• Design evaluation benchmarks and metrics (BLEU, ROUGE, custom checks) to audit hallucinations and domain alignment.
• Manage distributed DL training jobs across multi-GPU blocks and scale pipelines.
• Collaborate with MLOps to format model weights for cloud deployment and real-time inference serving.

Skills

Python

PyTorch

Transformers

PEFT

RLHF

DPO

CUDA

Tensor Parallelism

Education

Master’s or Ph.D. in CS/DS/Computational Linguistics

Tools

LoRA

QLoRA

Prefix Tuning

Prompt Tuning

GGUF

AWQ

GPTQ

Job description

NLP Data Scientist / LLM Fine-Tuning Specialist

• Employment Type: Contract
  
• Work Mode: Remote
  
• Location: Offshore
  
• Total Experience Required: 4 to 8 years
  
• Relevant Experience Required: 3+ years of dedicated natural language processing (NLP) and hands-on Large Language Model (LLM) fine-tuning experience
  
• Mandatory Certification: Google Cloud Certified Professional Machine Learning Engineer or AWS Certified Machine Learning - Specialty
  

Job Summary

We are seeking an experienced NLP Data Scientist / LLM Fine-Tuning Specialist to take ownership of our specialized open-source model optimization tracks. The ideal candidate will possess deep expertise in deep learning, dataset preparation, and parameter-efficient training methodologies to fine-tune foundational models for industry-specific terminology, domain-specific reasoning, and custom task execution.

Key Responsibilities

• Lead LLM fine-tuning initiatives, leveraging Parameter-Efficient Fine-Tuning techniques (PEFT) including LoRA, QLoRA, Prefix Tuning, and Prompt Tuning to optimize open-source architectures (e.g., Llama, Mistral).
  
• Curate, clean, and structure high-quality training datasets, implementing automated data deduplication, tokenization schemes, synthetic data generation pipelines, and human-in-the-loop validation frameworks.
  
• Implement advanced reinforcement learning alignment layers, configuring Reinforcement Learning from Human Feedback (RLHF) or Direct Preference Optimization (DPO) to enforce model safety, helpfulness, and tone guardrails.
  
• Optimize model footprint constraints and memory overhead, applying post-training quantization techniques (e.g., GGUF, AWQ, GPTQ) to minimize parameter degradation and compute budgets.
  
• Design rigorous evaluation benchmarks and metrics panels, executing automated validation tests (e.g., BLEU, ROUGE, custom verification matrices) to audit model hallucinations, factual accuracy, and domain alignment.
  
• Manage distributed deep learning training jobs, scaling pipeline configurations, tensor parallelism parameters, and gradient checkpointing scripts across multi-GPU compute blocks.
  
• Collaborate with MLOps infrastructure teams, formatting completed model weight checkpoints cleanly for scalable cloud deployment and real-time inference serving layers.
  

Requirements

• 4 to 8 years of core data science or advanced machine learning engineering experience, with 3+ dedicated years actively training, evaluation, and fine-tuning natural language processing systems.
  
• Expert-level technical mastery of Python, deep learning frameworks (PyTorch), transformer architectures (Hugging Face Transformers, Accelerate, PEFT), and vector calculations.
  
• Deep structural understanding of attention mechanisms, tokenization constraints, context window degradation behaviors, loss function optimization, and hardware compute limitations (CUDA).
  
• Mandatory certification: Professional ML Engineer or Specialty Machine Learning credential from a major cloud vendor (AWS/GCP).
  

Preferred Qualifications

• Master’s or Ph.D. in Computer Science, Data Science, Computational Linguistics, or an adjacent quantitative field with a research focus on neural network text models.
  
• Prior experience implementing custom embedding model structures or optimizing domain-specific classification layers inside constrained enterprise runtimes.