Aaron Baker

About Aaron Baker

I\'m a Chemical Engineer who became a Machine Learning Engineer and the chemistry didn\'t go to waste.Four years of CPG R&D taught me how to live inside messy, ambiguous problems: formulating a toothpaste that survives 24 months on a shelf, designing experiments where every variable is fighting the others, translating \"the flavor feels off\" into something measurable. It turns out that\'s most of the job in applied AI/ML too, the modeling is the easy part.Today I build production AI systems at Ashley (Factory Direct Enterprises). My current work spans an XGBoost classification pipeline standardizing a SKU catalog on Microsoft Fabric, a GPT-4o generative pipeline producing channel-specific product descriptions, an LLM-powered Copilot Studio chatbot serving ~100 internal users across the company, and a multi-source predictive returns model that\'s showing ~10x improvement over the existing baseline.Before Ashley, I spent four years at Colgate-Palmolive. First as a fragrance R&D technician, then through three R&D rotations (where I shipped an XGBoost mock-recovery classifier and a Python data-harmonization pipeline), and finally as the Hello brand\'s product development scientist, where I owned an end-to-end RAG tool (semantic chunking, all-MiniLM-L6-v2 embeddings, ChromaDB retrieval, OpenAI prompting, Pydantic validation) that codified tribal toothpaste-making knowledge for new scientists.On the side, I built and deployed GRASP which is a multi-service agentic web app (FastAPI + Celery + LangGraph + Postgres + React) that turns free-text meal descriptions into dependency-aware cooking schedules. Live on Railway and Cloudflare.(grasp.pages.dev)Stack: Python, SQL, Azure, GCP, LangGraph, RAG, vector DBs (Pinecone, pgvector, ChromaDB), MLflow, Docker, FastAPI, XGBoost, scikit-learn, TensorFlow.Currently exploring Machine Learning Engineer and AI Engineer roles in NYC or remote — especially in retail, CPG, or applied GenAI. Open to a conversation.

Experience

1. Ai Ml Engineer

   FDE LLC, d/b/a Ashley Northeast

   Oct 2025 — Present · Edison, NJ, US

   Built and deployed an XGBoost classification system on Microsoft Fabric (Lakehouse) that standardizes and enriches product attributes — including room category and product type — across the full SKU catalog, with data sourced via scheduled scrapes from manufacturer REST APIs.Engineered a GPT-4o-powered generative pipeline (Azure OpenAI) that produced channel-specific product descriptions for retail and e-commerce platforms, with a human-in-the-loop review layer for brand-voice and quality validation.Architected and launched an LLM-powered Copilot Studio chatbot serving ~100 internal users across Sales, Customer Service, Returns, and executive leadership; instrumented usage analytics to guide iteration.Scoped and delivered a multi-source predictive returns model integrating sales, customer, shipping, and product data with USPS API address validation. Pilot results indicate ~10x improvement in flagging preventable returns vs. the existing baseline.Contribute to Agile sprint planning and backlog design for current and future AI/ML initiatives; partner with Technology and Merchandising teams to translate workflow friction into structured technical roadmaps.

Education

• Rutgers University–New Brunswick

  Bachelor of Science - BS, Chemical Engineering

  2019 — 2023