AI / ML Engineer – LLMs & Self-Hosted AI at Navan · Tel Aviv

About the Role

We are looking for a Junior AI/ML Engineer to join our team building production AI systems for an advanced, agentic travel-support chatbot.

You will work with experienced ML and software engineers on open-source language models, fine-tuning, model evaluation, inference, and production integrations. This role offers hands-on exposure to the full AI lifecycle: preparing data, training models, evaluating quality, deploying experiments, and monitoring real-world performance.

What You'll Do

• Prepare, clean, and analyze datasets from production conversations and synthetic examples.
• Support supervised fine-tuning experiments for language models and classifiers.
• Experiment with techniques such as SFT, LoRA, and QLoRA.
• Compare base and fine-tuned models using offline evaluations and production shadow traffic.
• Evaluate model quality, including accuracy, routing, tool calling, hallucinations, and reliability.
• Integrate self-hosted models into agentic workflows and LLM-powered services.
• Help configure and test models using platforms such as SageMaker, vLLM, Impala, and Baseten.
• Analyze latency, throughput, token usage, failures, and cost.
• Build monitoring, logging, and evaluation tools for model experiments.
• Write tests for model integrations, routing, fallbacks, and shadow deployments.
• Debug issues across data pipelines, model endpoints, backend services, and agent workflows.
• Document experiments, datasets, model versions, and results.

What We're Looking For

• Degree, coursework, internship, or practical project experience in Computer Science, Machine Learning, Data Science, or a related field.
• Programming experience in Python, JavaScript, or TypeScript.
• Basic understanding of machine learning concepts, including training data, validation data, overfitting, and evaluation.
• Basic understanding of LLM concepts such as prompts, tokens, context windows, structured outputs, and tool calling.
• Familiarity with REST APIs, JSON, Git, and testing.
• Strong problem-solving and debugging skills.
• Ability to analyze data and investigate model-quality issues.
• Curiosity about open-source models, fine-tuning, and production AI systems.

Nice to Have

• Experience with PyTorch or Hugging Face.
• Hands-on experience with SFT, LoRA, QLoRA, embeddings, RAG, or text classification.
• Experience preparing conversational datasets or synthetic training data.
• Familiarity with Docker, AWS, SageMaker, vLLM, or GPU-based inference.
• Experience with Jest, Grafana, New Relic, or other observability tools.
• Familiarity with Node.js, React, Redis, or asynchronous programming.

Ideal Candidate

You are an early-career engineer who combines solid software fundamentals with a strong interest in AI. You enjoy experimenting, measuring results, and understanding why models succeed or fail. You are eager to learn, comfortable working across data and code, and excited to help turn research ideas and model experiments into reliable production capabilities.