AI Research Intern (PhD)

Job Description

AI Researcher Intern (PhD)

Salary: $25-$30 Hourly

About Akoncagua AI

Akoncagua AI builds domain-specialized AI systems for technically demanding industries. Our work includes LLMs, multimodal AI, agentic automation, RAG, evaluation, and production AI infrastructure. We are seeking AI Researcher Interns to help build secure, reliable AI agents and automated workflows.

What you'll work on

This role sits at the intersection of machine learning research, systems engineering, and product development. You will apply strong theoretical foundations in deep learning to build, train, adapt, evaluate, and optimize modern AI systems at scale.

\n
• Brainstorm, develop, and apply machine learning solutions to complex problems across a range of engineering industries.
\n
• Train, fine-tune, and evaluate large language models and vision-language models.
\n
• Research long-context learning, multimodal reasoning, post-training methods, and model reliability.
\n
• Analyze GPU memory usage and training throughput to improve the efficiency of large-scale training.
\n
• Develop data pipelines, training curricula, evaluation suites, and reproducible research workflows.
\n

Required qualifications

\n
• Currently pursuing a PhD in computer science, artificial intelligence, electrical engineering, applied mathematics, or a closely related field.
\n
• Completion of at least two years of PhD study before the internship begins.
\n
• Strong command of machine learning fundamentals, including probability, statistics, linear algebra, optimization, and experimental design.
\n
• Broad and rigorous understanding of deep learning across computer vision, natural language processing, and generative AI, including diffusion models and large language models.
\n
• Strong understanding of transformers, attention mechanisms, language modeling, tokenization, embeddings, and modern model architectures.
\n
• Hands-on experience training or fine-tuning models with PyTorch or a comparable framework.
\n
• Ability to design controlled experiments, establish appropriate baselines, select meaningful metrics, and analyze failure cases.
\n
• Experience independently solving open-ended research or engineering problems.
\n

Preferred qualifications

\n
• Publications at leading conferences such as NeurIPS, ICML, ICLR, ACL, EMNLP, CVPR, or MLSys, or evidence of comparable applied research contributions.
\n
• Research experience with LLMs, vision-language models, long-context models, or other large neural networks.
\n
• Experience fine-tuning and optimizing LLMs with tools such as Hugging Face Transformers, TRL, PEFT, or Unsloth.
\n
• Experience with distributed training frameworks such as PyTorch FSDP, DeepSpeed, or Megatron.
\n
• Understanding of distributed training and parallelism techniques, including data, tensor, pipeline, context, and sequence parallelism.
\n
• Experience with post-training methods such as instruction tuning, preference optimization, reinforcement learning, distillation, or synthetic data generation.
\n
• Prior work with financial platforms, databases, and transactions.
\n

How we work

We care about researchers and engineers who understand and can defend the design decisions in their code. AI coding tools are welcome here, but you should be able to explain why a system works the way it does, what evidence supports it, and where it breaks.

\n
• Our work connects model research with the systems required to run it. You will have room to explore new ideas, but you will also be expected to build working implementations, measure them carefully, and communicate their limitations clearly.
\n