AI Engineer – LLM Fine-Tuning, Chatbot Development, RAG & API Integration
Budget: ₹1,500 – ₹12,500 INR
I need a hands-on AI engineer who can own the full life-cycle of an intelligent, context-aware chatbot. The very first priority is fine-tuning a large language model so it understands our domain language and tone. Once the core model behaves correctly, we will wire it into a Retrieval-Augmented Generation pipeline that lets it answer from private documents stored in a vector database. Finally, the whole stack has to be exposed through a clean API or lightweight web interface that can be dropped into our existing product.
Here is what I expect at each stage:
• Model customisation – select an open-weights or API-based foundation model, prepare training data, and fine-tune it (PyTorch + Hugging Face/LoRA welcome).
• RAG layer – build embeddings, set up the retriever (FAISS, Pinecone, or similar), and connect it to the chat flow with LangChain or equivalent.
• Deployment – containerise the service (Docker/Kubernetes) and document the endpoints so my team can push it straight to our cloud.
• Testing & evaluation – include automatic metrics and live demos that show the assistant answering from the private corpus while refusing out-of-scope requests.
When you reply, attach a detailed project proposal describing-–in plain language–the tooling you would use, the milestones you foresee, and how you will measure success. I’m happy to discuss model size, inference budget, and data privacy requirements as soon as I see a plan that proves you have done this before or can learn fast.
Here is what I expect at each stage:
• Model customisation – select an open-weights or API-based foundation model, prepare training data, and fine-tune it (PyTorch + Hugging Face/LoRA welcome).
• RAG layer – build embeddings, set up the retriever (FAISS, Pinecone, or similar), and connect it to the chat flow with LangChain or equivalent.
• Deployment – containerise the service (Docker/Kubernetes) and document the endpoints so my team can push it straight to our cloud.
• Testing & evaluation – include automatic metrics and live demos that show the assistant answering from the private corpus while refusing out-of-scope requests.
When you reply, attach a detailed project proposal describing-–in plain language–the tooling you would use, the milestones you foresee, and how you will measure success. I’m happy to discuss model size, inference budget, and data privacy requirements as soon as I see a plan that proves you have done this before or can learn fast.