Founding Engineer - Full Stack / ML

Abacus; United States; Remote; USD 100,000 - 160,000; Full time

Job summary

An innovative company is seeking a Founding Engineer to refine and scale a cutting-edge document processing pipeline. This role involves taking ownership of both the frontend and backend, experimenting with LLMs, and making foundational architectural decisions. If you have experience building products from scratch and thrive in a dynamic environment, this is an exciting opportunity to make a significant impact in a rapidly growing startup. Join a team that values creativity and technical excellence in transforming document processing for CPA firms.

Qualifications

• Experience building products from the ground up (0 to 1).
• Strong background in frontend and backend development.

Responsibilities

• Experiment with and measure LLMs' performance for speed and accuracy.
• Take ownership of the Next.js frontend for document uploads.
• Evolve the Python/FastAPI backend and serverless pipelines.

Skills

Next.js

Python

FastAPI

OCR

LLMs

Serverless Architecture

Job description

We're looking for a Founding Engineer to play a key role in refining and scaling our existing document processing pipeline (think: ML tool for CPA firms). We've already shipped a working pipeline: a Next.js portal, a FastAPI backend, serverless orchestration (Modal), and a document processing engine built on OCR and LLMs. The Founding Engineer will help us scale it, clean it up, and make it exceptional.

Most important criteria: someone who's built something 0 to 1 (ideal signals: ex-CTO or engineer at a hyper-growth startup). I'm not opposed to candidates from large companies, as long as they can speak to something they've built.

What you'll be doing:

1. Experimenting with and measuring LLMs' performance for both speed and accuracy.

2. Take ownership of our Next.js frontend where users upload documents and view extracted data.

3. Own and evolve our Python/FastAPI backend, including serverless pipelines and async processing.

4. Get creative with OCR + LLM orchestration—combining Textract, layout models, and GPT-style models to pull structured data from chaotic PDFs.

5. Make foundational decisions across architecture, infra, and tooling.