Job Description & Details
This is a heavy-hitting GenAI and Data Science role where you won't just be playing with toy models; you'll be architecting and deploying real production-grade AI systems. If you are tired of standard CRUD apps and want to spend your days building RAG pipelines, agentic workflows, and LLM integrations, this contract gives you the sandbox to do it.
What You'll Actually Be Doing
You'll be knee-deep in designing AI-powered applications that solve actual enterprise bottlenecks. Your day-to-day will involve wrangling unstructured data, setting up robust Retrieval-Augmented Generation architectures, fine-tuning prompts, and stitching together LLMs like GPT or Claude with backend APIs and vector databases. You'll work closely with stakeholders to figure out how to translate vague business problems into scalable, reliable machine learning services that don't fall over the moment they hit production.
The Core Tech Stack
Python is non-negotiable here, and you need to be comfortable moving past basic scripting into building production-ready architectures. You'll need solid muscle memory with vector databases like Pinecone or ChromaDB, orchestration frameworks like LangChain or LlamaIndex, and traditional ML libraries such as PyTorch or TensorFlow. Cloud familiarity across AWS, Azure, or GCP is a must since these models need to live somewhere scalable.
Interview Expectations
Expect the hiring team to grill you on the failure modes of RAG and how you handle hallucination reduction in production environments. They'll likely ask you to whiteboard a complete architecture for an autonomous AI agent, focusing heavily on latency, cost optimization, and context window management. They want to see if you actually understand the underlying mechanics of vector search and embedding models, rather than just knowing how to import an API client.
Application Advice
Your resume needs to scream GenAI and production deployment, not just data analysis. Make sure you explicitly highlight any past projects involving LangChain, vector databases, and custom RAG pipelines to get past the initial ATS filters. Ditch the generic summaries and put quantifiable metrics front and center—show them the latency you dropped, the scale of data you processed, or the exact business value your AI models delivered.