This job has expired

This position was posted on October 10, 2025 and is likely no longer accepting applications. We've kept it here for historical reference. Check out the similar jobs below!

AI LLM Engineer

Trellis

Job Overview

Location

Indiana, USA

Job Type

Full-time

Full Job Description

📋 Description

• Architect the brain of America’s largest state-trial-court intelligence platform. You will design and own the end-to-end systems that turn hundreds of millions of unstructured legal documents into lightning-fast, semantically rich answers for attorneys nationwide.
• Own the LLM layer that powers Trellis’s flagship products. You will fine-tune open-source and proprietary models for legal-domain tasks such as claim extraction, motion summarization, judicial-behavior prediction, and conversational retrieval. Your models will directly influence how top-tier law firms litigate tomorrow’s cases.
• Build petabyte-scale data pipelines that ingest PDFs, OCR text, docket sheets, and minute orders; normalize them; and stream them into vector stores, graph databases, and feature stores with sub-second latency. You decide when to use SageMaker, Bedrock, or custom GPU orchestration to hit cost and speed targets.
• Create evaluation frameworks that measure factual accuracy, legal relevance, and citation integrity of LLM outputs. You will establish golden datasets, automated regression suites, and human-in-the-loop review workflows so that every model release raises the quality bar.
• Ship semantic-search experiences that let lawyers ask plain-English questions like “How did Judge Smith rule on motions to dismiss trade-secret claims in 2023?” and receive pinpointed case law, motion language, and outcome probabilities. You will tune embedding models, hybrid sparse-dense retrieval, and re-rankers to deliver Google-grade relevance on niche legal corpora.
• Collaborate daily with Product, Data Science, and Legal Content teams to translate attorney pain points into ML features. You will scope experiments, write technical specs, and present findings in sprint reviews where your impact is visible within weeks, not quarters.
• Mentor a growing team of ML and backend engineers through code reviews, design sessions, and pair programming. Your leadership will shape coding standards, MLOps practices, and the technical culture of a remote-first startup that tripled headcount last year.
• Champion responsible AI and privacy-by-design. You will implement guardrails that prevent disclosure of sealed records, detect bias in judicial-outcome predictions, and ensure SOC-2-grade audit trails for every model inference.
• Optimize cloud spend and inference latency as usage scales 10×. You will experiment with quantization, speculative decoding, and caching strategies to keep per-query costs fractions of a cent while maintaining 99.9 % uptime.
• Contribute to the open-source community. Whether releasing a legal-domain embedding model or publishing a benchmark dataset, you will amplify Trellis’s reputation as the technical leader in legal AI.

🎯 Requirements

• 7+ years of backend engineering experience, including 2–3 years shipping production features powered by Large Language Models
• 1–2 years fine-tuning LLMs (LoRA, RLHF, domain adaptation) for specialized, high-stakes use cases
• 1–2 years building or using LLM evaluation tooling—extractive/abstractive summarization metrics, retrieval-augmented generation benchmarks, or human-in-the-loop review systems
• 4–5 years deploying end-to-end ML pipelines on unstructured data; experience with SageMaker, Bedrock, or equivalent cloud ML stacks
• Python fluency and demonstrated ability to mentor other engineers; startup experience strongly preferred

🏖️ Benefits

• Remote-first culture with optional Los Angeles office space and flexible hours across US time zones
• 100 % employer-paid health, dental, and vision insurance for employees, spouses, and dependents
• 401(k) retirement plan with generous employer matching and immediate vesting
• Meaningful equity in a fast-growing legal-tech leader backed by top-tier investors
• Flexible vacation policy—take the time you need to recharge and avoid burnout

Skills & Technologies

Python

Remote

Ready to Apply?

Apply Externally

You will be redirected to an external site to apply.

Trellis

Visit Website

About Trellis

Trellis provides a legal research and analytics platform focused on U.S. state trial courts. It aggregates court filings, motions, rulings, dockets, judge and counsel data, and makes them searchable. Using AI tools, Trellis offers features like case assessment, motion drafting, and legal insights based on historical data and trends. It is meant to help litigators make more informed strategic decisions and improve efficiency in trial-court work.

View Company Profile