What does Aman Purwar do at Intuit?

Aman Purwar is a Staff Software Engineer on Intuit Go-To-Market Tech. He leads data-warehouse and real-time marketing data initiatives (including Snowflake adoption), builds AI-powered test automation (DC Nexus), and owns reliability for high-throughput visitor recognition services used for pre-authentication personalization.

What are Aman Purwar's technical focus areas?

His work spans distributed data platforms, Snowflake and streaming/batch pipelines, Kubernetes and cloud infrastructure, observability, and production GenAI patterns—including LLM evaluation, RAG-style prototypes, and AI-assisted software delivery.

Where did Aman Purwar study?

He earned an MS in Computer Science from Purdue University and a B.Tech in Computer Science and Engineering from IIT Kharagpur.

Staff Software Engineer · Intuit · Go-To-Market Tech

Aman Purwar

I build data platforms at scale, ship AI-powered automation in production, and lead reliability for high-traffic marketing systems—turning ambiguous cross-team problems into measurable customer and business outcomes.

Download résumé LinkedIn GitHub

About

I'm a Staff Software Engineer at Intuit on Go-To-Market Tech. Most of my recent work has been around building a Snowflake-backed marketing data warehouse at scale, shipping an AI-driven test automation platform, and making sure our visitor recognition service holds up at 5K TPS during peak season.

Before Intuit, I was a founding engineer at an LLM document-intelligence startup, worked on data connectors and security infra at Rockset (now part of OpenAI), and spent two years on market data systems at Tower Research Capital. I studied CS at IIT Kharagpur and then got my Master's at Purdue.

Outside of work — big F1 fan, I go on coastal Pacific hikes whenever I can, and I run a small financial research group on the side. Always happy to talk about any of the above.

Featured work

A few projects I'm proud of.

Marketing Data Warehouse

Snowflake · real-time pipelines · 15k–30k TPS

Led Intuit's first Snowflake adoption for marketing data, handling 15k–30k TPS. Cut end-to-end pipeline execution from ~50 minutes to 2 minutes and protected 20M+ user profiles from data quality incidents at peak season.

LLM Document Intelligence (Fulcrum AI)

Founding engineer · zero to one

Built the core LLM extraction and reconciliation stack from scratch as founding engineer— parallel document extraction with dynamic prompt injection, a revenue-critical reconciliation service, and an internal accuracy framework to keep models improving.

Random Access in Compressed Market Data

Tower Research Capital · systems engineering

Built an indexing layer that maps timestamps to positions inside gzip-compressed market data files, enabling direct seeks without full decompression. Brought index storage down from a projected 75 GB/day to ~2 GB.

Experience

Intuit

Staff Software Engineer, Go-To-Market Tech

Nov 2024–Present

Co-led the Marketing Data Warehouse initiative processing 15k–30k TPS, introducing Snowflake as Intuit's first adoption in this space; coordinated across 5+ teams and pioneered AI-assisted delivery (roughly 2× velocity) while protecting 20M+ user profiles from data-quality incidents.
Led DC Nexus, an AI-powered test automation platform from ideation to production—saving on the order of 1,000 hours/month of manual testing and shifting defect discovery earlier in the lifecycle.
Owned reliability and production readiness for a visitor-recognition service at 5K TPS (50ms latency budget), improving recognition from ~6% to ~72% for pre-auth personalization with zero peak-season incidents.

Fulcrum AI

Founding Engineer

Mar 2024–Nov 2024

Designed and shipped an asynchronous, LLM-powered parallel document extraction engine with dynamic prompt injection and few-shot examples—powering execution across policy, quote, and binder documents.
Built and deployed an LLM reconciliation service for a workflow representing a large share of company revenue; delivered a low-latency map–reduce style chat experience over document programs.
Developed Veriscriptum—an accuracy analysis and example-generation framework for extraction validation—improving internal accuracy and customer-facing confidence.

Rockset (acquired by OpenAI)

Software Developer, Data Connector Team

Jul 2023–Mar 2024

Implemented an AWS-backed encryption service for credentials and API keys with SOC 2–aligned controls.
Built an asynchronous parallel stats writer for API metrics persisted to FoundationDB, improving operational visibility.

Tower Research Capital

Software Developer II, Historical Market Data & Simulation

Jul 2019–Jul 2021

Prototyped random access for GZIP-compressed market data files.
Ran an Apache Spark pipeline archiving ~200 TB of market data, reducing storage overhead from small-file mirroring.
Upgraded a MongoDB metadata cluster (compression ~3×) and shipped an OAuth2 client API layer for metadata access.

Get in touch

Email works best. You can also find me on LinkedIn or GitHub.

Email ↗ LinkedIn ↗ GitHub ↗ Stack Overflow ↗