Python Developer - LLM/AI Agents

In this role, you will: Architect and implement modular LLM-driven agents that handle reasoning, planning, and tool usage. Operate and integrate LLMs across environments — from cloud-hosted models (e.g., OpenAI, Anthropic, Azure, Bedrock) to local/edge deployments (e.g., llama.cpp, vLLM, Ollama). Fine-tune and customize open-source LLMs using LoRA, QLoRA, PEFT, and other efficient tuning strategies. Monitor, evaluate, and debug agent performance using LangSmith, OpenInference, Promptfoo, or custom observability tools. Build retrieval-augmented generation (RAG) pipelines using vector stores like FAISS, Chroma or Pinecone. Orchestrate queries Collaborate with ML engineers, DevOps, and product teams to move prototypes into stable, production-ready systems. What you will bring: Strong experience in Python 3.x, with a focus on modular, well-tested code. Hands-on experience with LLMs in the cloud (OpenAI, Azure OpenAI, Claude, Bedrock, Hugging Face Inference Endpoints, etc.). Experience operating LLMs locally/on-device (e.g., using Ollama, llama.cpp, vLLM, GGUF/GPTQ models). Solid understanding of LLM agent frameworks (LangChain, CrewAI, AutoGPT) and prompt chaining. Practical experience fine-tuning models (e.g., using Hugging Face Transformers, PEFT, LoRA/QLoRA). Proficiency in using observability and evaluation tools (e.g., LangSmith, Promptfoo, TruLens) to trace, log, and improve performance. Experience with embeddings, vector databases, and contextual memory systems. Nice to Have: Experience deploying LLMs to Kubernetes, serverless platforms, or edge devices (e.g., Jetson, mobile, Raspberry Pi). Familiarity with quantization and optimization for low-latency inference. Built or contributed to multi-agent systems or long-context memory chains. Experience with model monitoring, alerting, and auto-evaluation pipelines in production.