AI Engineering Intern (LLM)

Summary: Veolia is the top-ranked environmental company in the United States, offering a full spectrum of water, waste, and energy management services. The AI Engineering intern will support the development and implementation of an AI-powered deep research agent, gaining hands-on experience with large language models and cloud infrastructure. Responsibilities: - Understanding and working with commercial/proprietary LLMs such as Gemini( Google), GPT(OpenAI) and Claude Sonnet (Anthropic)for high performance, large context, and multimodal tasks - Familiarity with open-source/self-hosted LLMs like Llama from Meta and Mixtral from (Mistral AI) - Requirements Gathering: Using Confluence for documentation and collaboration - Architecture Design: Creating system diagrams and workflows with Lucidchart - Prototyping: Designing UI/UX prototypes in Figma - Project Management: Tracking tasks and progress in Jira - Data Preparation & Management: Cleaning, transforming, and organizing data for use in AI/ML workflows - Using LangChain or LlamaIndex for orchestrating LLM applications - Building multi-agent systems with Semantic Kernel, CrewAI, and LangGraph - Managing and optimizing prompts with LangSmith - Implementing semantic search and retrieval using Vertex AI Vector DBs - Developing RESTful APIs with FastAPI (Python) - Integrating asynchronous communication with Apache Kafka and Redis Streams - Building user interfaces with React or Angular - Utilizing Material-UI for consistent, modern UI elements - Writing and debugging code in VS Code - Leveraging GitHub Copilot and Cursor for code suggestions and productivity - Managing code with GitHub, or GitLab - Ensuring code quality and standards with SonarQube, ESLint, and Pylint - Using Vertex AI Tuning for model customization - Training and experimenting with models in PyTorch, TensorFlow, or JAX - Applying parameter-efficient fine-tuning (PEFT) methods like LoRA and QLoRA - Generating synthetic data - Assessing models with HELM, lm-evaluation-harness, and custom benchmarks - Using RAGAS, and DeepEval for LLM evaluation; LangSmith Evaluators for prompt testing; hallucination detection - Packaging applications with Docker - Managing containers at scale with Kubernetes and Google GKE - Using Google Cloud Platform (GCP) services such as Vertex AI for ML, GKE for Kubernetes, Cloud Run for serverless deployment, and Cloud Functions for event-driven tasks - Monitoring LLM performance and usage with LangSmith and Weights & Biases - Monitoring and optimizing costs with OpenMeter and custom dashboards - Setting up continuous evaluation pipelines to ensure model quality and reliability Required Qualifications: - Working towards a PhD degree in AI/ML/Computer Science - 3.8 Cumulative G.P.A required - Strong communication skills, including written, verbal, listening, presentation and facilitation skills - Demonstrated ability to build collaborative relationships Required Skills: Large Language Models, Cloud Infrastructure, API Development, Model Fine-tuning Important Skills: Data Preparation, Frontend Development, Backend Development, Version Control Nice-to-Have Skills: Collaborative Relationships Benefits: Medical and basic life insurance coverage is available to temporary employees scheduled to work 20 or more hours per week immediately following 60 days of service., Veolia observes 11 holidays.