AI Engineer (LLM Fine-Tuning)

Location: Remote in Argentina, Brazil, Peru, Colombia, Costa Rica, or Mexico; U.S. timezone overlap required

Company

hirify.global is presenting a role for medxprts.ai, which is building an AI-powered platform for legal and healthcare applications using LLMs, agentic workflows, and automation.

What you will do

• Design, run, and ship LLM fine-tuning experiments using LoRA, QLoRA, and full SFT on open-weight models such as Llama, Mistral, and Qwen.
• Build training infrastructure with PyTorch, Hugging Face tooling, multi-GPU orchestration, and AWS or GCP.
• Engineer datasets from unstructured documents, including medical and legal records, with deduplication, filtering, contamination checks, and train/evaluation splits.
• Create evaluation harnesses with held-out test sets, human-calibrated LLM judges, regression tests, and model-drift monitoring.
• Implement DPO/RLHF-style feedback workflows using expert corrections and integrate them into retraining pipelines.
• Collaborate with backend and frontend engineers on model integration, inference optimization, production deployments, debugging, and release processes.

Requirements

• Hands-on experience fine-tuning open-weight LLMs with LoRA, QLoRA, and full SFT, including at least one shipped example.
• Experience with PyTorch, Transformers, PEFT, TRL or Axolotl, DeepSpeed or FSDP, and training workloads on AWS or GCP.
• Practical expertise in dataset engineering, evaluation harnesses, preference learning, DPO/RLHF-style workflows, and expert-feedback integration.
• Strong software engineering fundamentals, including Git, pull requests, testing, debugging, deployment, APIs, databases, and service integration.
• Good English communication skills and availability to overlap with U.S. working hours.
• Applicants must be located in Argentina, Brazil, Peru, Colombia, Costa Rica, or Mexico.

Nice to have

• Experience with long-context strategies, retrieval-aware training, context extension, or RAG for very large documents.
• Model deployment and inference optimization using quantization, vLLM, TGI, batching, and throughput tuning.
• Experience with vector databases, advanced RAG pipelines, MCPs, or workflow automation tools.
• Knowledge of Docker, CI/CD, Kubernetes, EKS, or serverless infrastructure.
• Background in healthcare, legaltech, HIPAA-aware processes, or other regulated and data-sensitive environments.

Culture & Benefits

• Fully remote work.
• High ownership and autonomy.
• Opportunity to build production AI products for legal and healthcare applications.
• Performance-based incentives and outcome-driven bonuses.
• Potential transition from part-time to a long-term full-time role.