The posting
Exodigo is the leading underground mapping solution for non-intrusive discovery. Our platforms combine multi-sensor fusion, 3D imaging, and AI technologies to create complete, accurate underground maps that enable confident decision-making for customers across the built world. We transform the project lifecycle for our customers, who include key community stakeholders in the utilities, transportation and government sectors.
We are experiencing sky-rocketing growth and closed a historically large $96M Series B round in July of 2025.
Job description
Our Algorithms group is a highly multidisciplinary group at the core of our data processing and detection capabilities, integrating physics, signal processing, computer vision, classical algorithms, and AI models to meet our unique requirements. Our problems are grounded in the physical world: sensor data, engineering drawings, coordinate systems, and the messy reality of infrastructure that was built decades ago and documented inconsistently ever since.
Job Description
We are looking for an AI Algorithms Engineer to join our Computer Vision team and bring LLMs and agentic AI into problems that are anything but generic. This role is for someone who wants to keep their algorithmic craft and add to it — not trade it in. The interesting work here sits exactly where modern AI meets geometry, vision, and physical measurement, and it needs someone who is genuinely strong on both sides of that line.
The work is varied by nature. It can mean finding and making sense of engineering documents that were written by people for people, anchoring what they contain to real-world coordinates, reconciling spatial data from partial and conflicting sources, or building tools and agents inside the professional software our experts work in — platforms such as Civil 3D and ArcGIS. These are separate problems, not stages of one pipeline, and new ones appear regularly.
What ties them together is judgment. Some of these problems are a natural fit for LLMs and agents; others are solved far better with geometry, classical computer vision, or a well-chosen heuristic — and many need both. Knowing which is which, and being honest about it, is the single most valuable thing you will bring to this team.
Our problems also rarely arrive fully specified. You will often start from a one-line description and an unclear success criterion, in an environment where priorities shift. We are looking for someone who is comfortable in that space — who can decompose an amorphous problem, identify what information is missing, and then go get it: talk to domain experts, dig through data, run a quick experiment, and come back with a sharper definition of the problem than the one they were handed.
Key Responsibilities
- Own problems end to end — from an ambiguous initial ask, through problem definition and prototyping, to a system that runs reliably in production.
- Design and build LLM-powered systems and agentic workflows that automate complex, multi-step tasks involving spatial data, engineering documents, and professional software.
- Combine modern AI with classical approaches — including geometry, computer vision, heuristics, and optimization — and make deliberate decisions about which approach best fits each part of the problem.
- Work directly with mapping experts, surveyors, and algorithm researchers to understand real-world workflows and turn that understanding into tools people actually use.
- Evaluate and integrate new foundation models, frameworks, and techniques, and design evaluation methodologies that determine whether these systems are genuinely reliable on our data.
Requirements
- 4+ years of hands-on experience building algorithmic, machine learning, or AI systems that were deployed and used in production.
- B.Sc. in Computer Science, Computer Engineering, Electrical Engineering, Mathematics, or a similar field.
- A strong algorithmic background is a hard requirement. We are looking for real depth in at least one of the following areas: computer vision, data science, NLP, deep learning, geometry and 3D data, or signal processing. This is the part of the profile we are strict about.
- Genuine interest in LLMs and agentic AI, with a practical understanding of how they work and where they fail. You do not need production experience with agents — a side project, prototype, or serious personal exploration is enough, as long as you can walk us through what you built and why. What matters is that you want this to become a real part of your craft.
- Comfort working in a dynamic, loosely defined environment. Problems here arrive amorphous and priorities shift. You should be able to take a vague problem, break it into something tractable, identify what information is missing, and go get it — by talking to domain experts, digging through data, or running a quick experiment.
- Strong proficiency in Python and solid software engineering practices, including clean, maintainable code and robust systems.
Preferred Qualifications
- Experience with multimodal and vision-language models, especially for document understanding, drawings, or diagrams.
- M.Sc. or Ph.D. in a relevant field.
- Experience with geospatial data, GIS, CAD, or coordinate reference systems.
- Experience automating or extending professional software through its APIs or plugin ecosystem.
- Familiarity with cloud development environments, particularly AWS.
A Note on This Role
This is deliberately a broad role, and we know that very few people cover all of it. If you have real algorithmic depth and are seriously curious about LLMs and agents — even if your experience with them so far is a side project rather than years of production work — we would like to hear from you. Tell us what you are strong at and what you want to grow into.
