Build Agentic AI In-House
Publicada el 2026-07-17
Descripción de la oferta
I’m kicking off an initiative to bring agentic, retrieval-augmented generative AI (RAG + LLMs) fully inside our organisation. The goal is to move beyond simple prompt engineering and create production-ready systems that can reason, act autonomously, and serve domain-specific content with low latency. You’ll help across the whole stack: shaping data pipelines that feed vector stores, fine-tuning or training large language models for our private datasets, optimising them for speed and cost, and then wiring everything into the services we already run. Smooth integration matters as much as model quality; the finished solution must expose clear REST or gRPC endpoints so our back-end and front-end teams can consume it without friction. Key deliverables • Scalable data-ingestion and vector-store pipeline powering a RAG workflow • Fine-tuned / optimised LLM (or set of models) meeting defined accuracy and latency targets • Agent framework that enables multi-step tool calls and decision making • Clean, documented codebase in a shared Git repo plus deployment scripts for our cloud environment • Integration layer and example clients proving end-to-end functionality in staging Acceptance criteria • Reproducible training pipeline and inference setup • Latency and accuracy benchmarks achieved on agreed test set • All components deploy with one-click (CI/CD or IaC) and pass automated tests Tech we already use: Python, LangChain, PyTorch, AWS (EKS & SageMaker). Open to your preferred libraries if they mesh with these. If you have experience delivering similar agentic or RAG systems—from data engineering through to production integration—I’d love to collaborate and get this live quickly.
Skills
- REST APIs
- Machine Learning
- AWS
- Python
- RAG
- Vector Databases
- Data Engineering
- PyTorch
- Prompt Engineering
- LangChain
- CI/CD
- gRPC
- Git
Fuente original: freelancer