About Lighthouz AI

Lighthouz AI is automating the back office of freight finance with freight-native AI agents. We help freight brokers, 3PLs, and factoring companies process invoices, rate confirmations, and PoDs in seconds—not hours—by replacing manual audits and brittle RPA with intelligent automation.

Our platform handles real-world document chaos—scanned and handwritten paperwork, NOAs, BOLs, emails, and portal logins—executing complex workflows automatically. The result: faster payments, fewer disputes, and 10x operational throughput.

We’re a Y Combinator S24 company founded by a team with deep experience across AI, supply chain, and enterprise systems (Google, Georgia Tech, Progressive). At Lighthouz, we’re not just streamlining freight finance—we’re rebuilding it from the ground up.

About the Role

We’re seeking a passionate AI Engineer (Document vision) to design, build, and maintain intelligent agents that will automate and transform freight workflows. You’ll work on the two core components of our AI agents – first, the core perception systems that extract structured insights from messy, real-world freight documents—handwritten, scanned, distorted, or multi-page – and second, our AI agents for email and voice communications between freight entities. You will push the boundaries of creative prompt engineering to solve real-world problems at scale., fine-tuning LLMs, building large-scale document classification and entity extraction models, communication understanding, intent classification – your code will be at the heart of automating financial decision-making in freight.

You’ll collaborate closely with the backend and product teams to bring AI models to life in production environments and continuously improve performance in the wild.

What You’ll Do

Architect, implement, and deploy AI agents for email and phone communications between freight accounting parties (payer/payee), leveraging language and vision LLMs for automation and analysis.
Design and refine high-impact prompts, templates, and evaluation harnesses to ensure robust, reliable agent behavior.
Build scalable pipelines for preprocessing, training, inference, and feedback loops, including evaluation and integration of VLMs.
Monitor and diagnose agent performance in production, rapidly addressing failures and refining prompts, workflows, and models.
Create and maintain high-quality training, evaluation, and test datasets.
Enhance the AI stack through creative prompting, fine-tuning, and continuous iteration.
Productionize models within Lighthouz’s intelligent automation platform.
Collaborate with product and engineering teams to integrate AI outputs into document, email, and voice workflows, delivering polished, production-ready solutions.
Continuously improve model performance in real-world conditions.

What We’re Looking For

4+ years in ML/AI roles, ideally in document AI
Deep learning expertise with PyTorch, TensorFlow
Prompting wizardry — skilled at crafting precise, reliable prompts for VLMs & LLMs, translating complex tasks into actionable instructions
Experience fine-tuning VLMs
Strong knowledge of OpenAI, Claude, and other agentic toolkits
Hands-on with OCR, visual transformers, multimodal models
Background in conversational AI, voice AI, NLP research, and LLM training
Proven track record of training & deploying models to production
Problem-solver & builder mindset — fast to prototype, faster to iterate
Comfortable with ambiguity and evolving datasets

Nice to Have

Familiarity with freight, logistics, or fintech workflows
Experience with AWS, Azure, or GCP-based ML infrastructure
Exposure to RAG pipelines, foundation models, or vector search systems
Knowledge of document layout understanding (e.g., Donut, LayoutLM, PubLayNet)
Background in building secure, production-grade ML services

What We Offer

💰 Competitive salary

🌎 Fully remote. For US – no sponsorship is provided.

🛠️ High ownership, zero bureaucracy—help shape our AI stack from day one

🚀 Work on impactful real-world problems that blend AI and automation at scale