Artificial Intelligence Intern

About the Role

Brainy Neurals is a founder-led AI engineering company seeking interns to work on production AI for clients across 15+ industries. This internship focuses on deep understanding of AI models, going beyond API calls and prompt writing. It explicitly excludes data analysis, dashboards, Excel/Power BI reporting, and Kaggle-style tabular modeling.

Key Responsibilities

• LLM Engineering: Fine-tuning open-weight models with LoRA and QLoRA, and building evaluation sets to assess fine-tune effectiveness.
• Edge AI: Quantizing and converting models with TensorRT and ONNX, and measuring latency and accuracy on NVIDIA Jetson hardware.
• Agentic AI: Building agents that plan, call tools, recover from failed steps, and provide debuggable traces.
• Documentation AI: Parsing PDFs and scanned pages, and generating GMP documents for pharma.

Requirements & Tech Stack

• Python and hands-on PyTorch, including experience writing a training loop from scratch.
• At least one project involving model training or fine-tuning with measured results, with accompanying code on GitHub.
• Ability to interpret research papers or model cards and translate them into working code.
• Hands-on experience with Hugging Face Transformers, vLLM, llama.cpp, or Ollama (Good to have).
• Experience deploying models with TensorRT or ONNX Runtime on edge devices like Jetson boards (Good to have).
• Experience building an agent using LangGraph or plain Python for multi-step tasks (Good to have).
• Work on OCR or PDF parsing where table and scanned page extraction was required (Good to have).

Technical Concepts to Explain

Candidates should be able to explain the following concepts at a whiteboard, without notes, in plain words:

• How self-attention works and its compute cost.
• Differences between decoder-only models (e.g., Llama) and encoder-only models (e.g., BERT), and their respective use cases.
• When to choose full fine-tuning, LoRA, QLoRA, or DPO over prompting and retrieval, with justifications.
• The impact of KV cache, context length, batch size, and INT8 quantization on latency and memory on GPUs.
• The end-to-end process of a training step, including forward pass, loss, backpropagation, optimizer update, and identifying overfitting from loss curves.

Qualifications

• B.E., B.Tech, M.E., or M.Tech in CS, IT, EC, or AI/ML, completed in 2026 or due in 2027.

Internship Details

• Duration: Six months, full time.
• Location: Ahmedabad office only (Work from office).
• Work involves live client projects and company products, with senior engineer review of all pull requests.
• A monthly stipend will be provided.
• Final-year students require a college letter confirming full-time availability for the internship duration.