AI Engineer Intern Full Time

Elimai.ai is an AI-native software company building next-generation intelligent platforms formanufacturing, healthcare, and enterprise operations. We specialize in building intelligentagentic workflows. We combine cutting-edge LLMs, specialized language models (SLMs), andcustom agents to solve real operational problems for mid-market businesses in India.Our tech stack spans RAG systems, MCP servers, full-stack ML, and production-gradeinference pipelines. We ship, iterate, and learn in public. No slow bureaucracy.The RoleAs an AI Engineer Intern, you’ll contribute to building production AI systems that processmillions of extracted structured/ unstructured data from unstructured documents andorchestrate multi-step agentic workflows. This isn’t a “learn Python” internship—it’s hands-onengineering on real products with real users.What You’ll Build• Agentic workflows: Design and implement multi-turn agent loops that parse emails, extract intent, fetch context from APIs, and generate structured responses using varied LLMs• MCP servers: Build Model Context Protocol integrations to connect AI agents with internal databases, ERPs (NetSuite, SAP), and custom tools. These become moats for our product.• LLM/SLM fine-tuning: Fine-tune small language models for domain-specific tasks (invoice extraction, PO parsing, buyer intent classification) on labeled datasets. Measure BLEU/F1 gains vs. base models.•RAG intelligence layers : Build retrieval pipelines over RabbitMQ emails and documentstores. Implement vector embeddings (text-embedding-3-small), vector DBs (Pinecone), and semantic search.• Full-stack ML: Train, evaluate, and deploy models end-to-end. Own the notebook → experiment tracking (Weights & Biases)→ dataset versioning (DVC)→ inference server (FastAPI)→ monitoring pipeline.Required Skills• Python expertise.LLM fundamentals: Understand prompt engineering, token budgeting,context windows, and why you use different models for different tasks.• ML basics: Training/validation/test splits, evaluation metrics (F1, precision, recall),overfitting, cross-validation. Can read a paper and implement it.• SQL & databases: Write efficient queries, understand indexing, normalization.Comfortable with PostgreSQL or similar.• Git & DevOps basics: Version control, Docker basics, CI/CD pipelines. Understand whywe containerize inference.Curiosity > credentials: Strong fundamentals beat fancy resume items. We care thatyou can debug, think in systems, and ship.Good-to-Have• Experience with LangChain, LlamaIndex, or similar agent frameworks.• Fine-tuning experience: LoRA, QLoRA, or custom training runs on HuggingFace models.• RAG at scale: Built or debugged RAG pipelines in production.• MCP experience: Built or integrated Model Context Protocol servers. Experienceworking in VLMs is a plus.What We Offer• ₹10,000-14,000/month stipend: Hybrid internship in Coimbatore. Open for Initial Remote work but should move to Coimbatore if things are permanent.• Real code, real impact: Your work goes to production. You’ll see users, revenue, and feedback.• Full-stack ownership: From idea → training → deployment → monitoring. Own your responsibilities in the project end-to-end.• Direct mentorship: No corporate gatekeeping.• Potential extension: Strong performers can transition to a full-time role or continued contract work post-internship.Day-to-Day Responsibilities• Build and ship features: Implement a component (agent, MCP server, fine-tuned model)from spec to production in 1-2 week sprints.• Debug production issues: Investigate why an email pipeline broke, a model’s accuracydropped, or an API returned garbage.• Read & understand papers: Stay current. We share research on in-context learning,retrieval optimization, and agent architectures.• Participate in code reviews: Get feedback, give feedback. Shipping quality code is non-negotiable.• Experiment & measure: Run A/B tests on prompt templates, model choices, andretrieval strategies. Document findings.