MTS - Engineering (Data Infrastructure)

Collinear AI, Inc.; San Francisco (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Collinear AI, Inc. in San Francisco is seeking a Member of Technical Staff - Engineering (Data Infrastructure) to own the systems that turn large, real-world datasets into data Collinear can build on. Our environments span terabytes across archives, spreadsheets, email, PDFs, and scanned documents.

This hands-on role sits at the intersection of algorithms, distributed systems, and data quality. You will define scalable solutions for linking related records, fast search, and data cleansing, while

Qualifications

• 5+ years of experience building data-intensive systems in production.
• Experience with parallel processing of large datasets using Spark, Ray or Dask.
• Strong command of graph algorithms, and transforming large amounts of data.
• Experience with efficient string matching and pattern search at scale.
• Ability to trade off accuracy, speed, and cost, and explain decisions clearly.

Responsibilities

• Build pipelines that process multi-terabyte datasets in parallel across archives, spreadsheets, email, PDFs, and scanned documents.
• Design graph-based systems that link related records across millions of files.
• Build fast string and pattern search over large, heterogeneous datasets.
• Profile bottlenecks and decide how to split work that doesn't parallelize neatly.
• Transform data for downstream use, including replacing sensitive fields across files.
• Define data quality metrics and build review tools to catch and fix errors without full reprocessing.

Skills

Data-intensive systems

Distributed systems

Graph algorithms

String matching

Data quality

Data pipelines

Parallel processing

Systems programming (Rust/Go/C++)

Tools

Apache Spark

Ray

Dask

Job description

About the role

As a Member of Technical Staff - Engineering (Data Infrastructure), you will own the systems that turn large, real-world datasets into data Collinear can build on. Our environments are grounded in terabytes of data, spread across archives, spreadsheets, email, PDFs, and scanned documents. The pace at which we process these determines how fast we deliver to frontier labs.

This is a hands-on role at the intersection of algorithms, distributed systems, and data quality. Many of our hardest problems, such as linking related records across millions of files, don't split up neatly, and you will define how we solve them at scale.

What you'll do

• Build pipelines that process multi-terabyte datasets in parallel across archives, spreadsheets, email, PDFs, and scanned documents

• Design graph-based systems that link related records, such as the same person or company appearing across millions of files

• Build fast string and pattern search over large, heterogeneous datasets

• Profile and remove bottlenecks, and decide how to split work that doesn't parallelize neatly

• Transform data for downstream use, including consistently replacing sensitive fields across files

• Define how we measure data quality, and build review tools so the team can catch and fix errors without reprocessing everything

• Assess new datasets and filter out low-quality data before it reaches our environments

About you

• You have 5+ years of experience building data-intensive systems in production

• You have processed large datasets in parallel with frameworks such as Apache Spark, Ray, or Dask, and know when to design your own

• You have a strong command of graph algorithms, and experience using them to transform large amounts of data

• You have built efficient string matching and pattern search at scale, such as fuzzy matching or indexing

• You make sound tradeoffs between accuracy, speed, and cost, and can explain them clearly

Nice to have

• Experience with entity resolution or record linkage

• Experience with NLP or LLM-based information extraction

• OCR or document processing experience, including poor scans and handwriting

• Experience in a systems language such as Rust, C++, or Go

• Experience with regulated or sensitive data, such as financial or healthcare records