// hi, my name is
Data engineer by trade. I build pipelines that don't break at 2 AM, and RAG systems that actually retrieve the right context. Turning raw chaos into clean, queryable truth.
I'm Arsh Sahay — a data engineer obsessed with turning raw data into actionable outcomes. I design scalable architectures, ship end-to-end pipelines, and increasingly, build agentic AI systems that actually do useful work.
I work comfortably across AWS and Azure — Cosmos DB, Blob Storage, SQS, Redis, the lot. My happy place is a clean DAG, a self-hosted service on a Raspberry Pi, and a proof that the thing works at 2 AM without paging anyone.
I'm also a Python geek, an ML enthusiast, and yes — an ETL wizard and pipeline builder. When I'm not refactoring infra, I'm fine-tuning an open-source LLM to turn the lights off.
# identity
name: "Arsh Sahay"
role: "Data Engineer"
location: "remote-first"
mode: "always_on"
# self_described_as
- ml_enthusiast
- python_geek
- etl_wizard
- pipeline_builder
# currently
- data_engineer: "@ Digihumans"
- self_hosting: "rpi5 + docker"
- side_quests: "agentic ai"An end-to-end EDA pipeline over the NYC 311 service-requests dataset. Cleans, engineers features, and visualizes complaint patterns and resolution times across boroughs. Includes a missing-value analysis with a 50%+ threshold and full geographic breakdowns.
An LLM-powered agentic assistant that interprets natural language weather queries and autonomously calls a weather API. Backend self-hosted on a Raspberry Pi 5 inside Docker; Cloudflare handles HTTPS and tunneling. Frontend built with Kiro.
An open-source LLM with custom tool-calling. Interprets natural-language commands and autonomously controls smart lights through Home Assistant. Fully self-hosted for privacy. Roadmap: fine-tuning, voice input, and room-level context awareness.
I work at the intersection of data engineering and intelligent retrieval. As a Data Engineer, I build scalable pipelines that ingest, transform, and structure raw data into usable products. At Digihumans, I designed cloud-native workflows on AWS and Azure, built document processing pipelines for PDFs, DOCX, PPT, and PPTX files, and developed a RAG system that preprocesses documents into markdown, applies intelligent chunking, enriches content with metadata, and uses embeddings for efficient retrieval. I also implemented orchestration and automation using Redis, AWS SQS, Terraform, Ansible, and Jenkins to keep the system reliable and scalable.
Built ETL pipelines to extract data from multiple document formats (PDF, DOCX, PPTX). Wrote transformation logic to unify them into a clean markdown format, then loaded them into Azure Blob Storage and Cosmos DB. Automated document processing and web scraping workflows in Python.
Started with a CS foundation, a curious mind, and a Jupyter notebook that didn't crash. The rest is git history.
$ _
Initializing handshake...
Routing message...
Ready to receive input.