- Company Name
- Veolia
- Job Title
- Intern - AI Engineering SLM
- Job Description
-
**Job Title**
Intern – AI Engineering SLM
**Role Summary**
Support the design, build, test, deploy, and monitor of small language model (SLM) applications for enterprise use. Apply cutting‑edge lightweight models (e.g., Phi‑3, Llama, Mistral, Gemma, TinyLlama) and associated tooling (LangChain, vLLM, ChromaDB, FastAPI). Deliver production‑ready solutions on edge, local, or cloud platforms while collaborating with senior AI engineers and stakeholders.
**Expactations**
- Complete end‑to‑end SLM application development within a 12‑week rotation.
- Demonstrate proficiency with model selection, fine‑tuning, prompt engineering, and RAG pipelines.
- Produce clean, version‑controlled code and document experiments.
- Participate in sprint reviews and knowledge‑sharing sessions.
**Key Responsibilities**
1. Gather functional requirements and design UI/UX workflows in Figma.
2. Explore, clean, and analyze data using Jupyter, pandas, and notebooks.
3. Select and download pre‑trained SLMs from Hugging Face Model Hub; benchmark performance.
4. Develop application logic with LangChain or Lanngraph, orchestrating workflows.
5. Fine‑tune models with PEFT (LoRA/QLoRA) via Hugging Face Transformers.
6. Implement semantic search and RAG with ChromaDB and evaluate outputs.
7. Build RESTful APIs using FastAPI (Python) or Express.js (Node.js).
8. Containerize workloads with Docker; release on Kubernetes or Docker Swarm.
9. Create and run unit, model, and load tests (pytest, RAGAS, DeepEval, Locust).
10. Log experiments with MLflow/Weights & Biases and track data with DVC.
11. Monitor and observe application performance through LangSmith and APIGateway tools.
12. Deploy on chosen infrastructure (on‑premise, GCP Vertex AI, or hybrid edge).
**Required Skills**
- Programming: Python, Git, VS Code.
- AI/ML: Familiarity with small language models, Hugging Face, LangChain, vLLM.
- Data handling: pandas, Jupyter.
- Model fine‑tuning: LoRA, QLoRA, Hugging Face Transformers.
- Prompt engineering: LangSmith, PromptLayer.
- API development: FastAPI or Express.js.
- Testing frameworks: pytest, unittest, RAGAS, DeepEval, Locust.
- Containerization: Docker, Docker Compose.
- Orchestration: Kubernetes (K8s), Docker Swarm.
- CI/CD: GitHub Actions, GitLab CI.
- Experiment tracking: MLflow, Weights & Biases.
- Version control: Git, DVC.
- Cloud basics: GCP Vertex AI or equivalent.
- Communication: strong written and verbal skills.
**Required Education & Certifications**
- Pursuing a Ph.D. (or equivalent advanced degree) in AI/ML/Computer Science.
- Minimum cumulative GPA of 3.8.
- No additional certifications required, but knowledge of cloud provider (GCP, AWS, Azure) is advantageous.