Sr. AI Engineer - RAG/LLM Systems

Overview

About Us
We are NymbleUp - we build workforce management software that actually works. Our platform helps retail and restaurant chains figure out how many people they need, when they need them, and creates schedules that don't make managers want to pull their hair out.

We've got demand forecasting running on Prophet, scheduling algorithms we've been iterating on for years, and a Django backend that processes millions of shift calculations. Now we want to make it smarter - and that's where you come in.

*What You'll Actually Do *

● Build a conversational layer on top of our scheduling and forecasting engines. Right now, if a location manager wants to "optimize for evening rush" or "account for a nearby event", they click through 15 screens. We want them to just... say it.

● Design and implement RAG pipelines that understand our domain - labor laws, shift rules, coverage requirements, historical patterns. Not generic chatbot stuff.

● Hook LLMs into our existing Python services so they can actually trigger schedule generation, adjust forecasts, and modify configurations - not just answer questions about them.

● Figure out multi-agent orchestration for complex requests like "replan next week considering any employee is on leave, there's a big event on Saturday, and we're short on evening trained staff."

● Work with messy real-world data. Our clients have 10 years of POS data, attendance records, and about a hundred special rules per store. Your models need to make sense of this without hallucinating.

*What We're Looking For *

Must have:

● 4+ years building production ML systems (not just notebooks)

● Hands-on experience with RAG architectures - embeddings, vector stores, retrieval strategies, chunking that actually matters

● Worked with LLMs beyond API calls - fine-tuning, prompt engineering that scales, dealing with context limits

● Python is your primary language. Django experience is a plus but not required

● You've debugged why an LLM gave a wrong answer and fixed it without just "trying different prompts"

*Really helpful: *

● Built multi-agent systems or agentic workflows (LangGraph, AutoGen, CrewAI, or your own thing)

● Experience in retail operations domain - retail, logistics, hospitality will be an advantage
● Dealt with time-series data and forecasting

● You know when NOT to use LLM

*Tech You'll Work With *

● Backend: Python, Django, Celery, PostgreSQL

● AI/ML: LangChain/LlamaIndex (or you tell us what's better), vector DBs, OpenAI/Claude/open-source LLMs

● Forecasting: Prophet, xgboost and custom time-series models

● Infra: GCP, AWS

*Why NymbleUp *

● You'll work on AI that affects real shift workers - not just another chatbot

● Small team, big impact. Your code runs in 1000+ locations across India, SEA and Middle East
● We ship fast. If your idea works, it's in production next sprint

● No "AI theater" - we care about results, not demos

We're not looking for someone who knows all the buzzwords. We're looking for someone who can make our scheduling engine talk back.