AI Engineer

At Esper, we are building the sensor infrastructure to measure the physical world. Our hyperspectral imaging systems capture data beyond the visible spectrum, providing critical insights for mineral exploration and environmental monitoring. As we scale our infrastructure and push for stronger outcomes from our sensing stack, we're hiring an AI Engineer to join our team and help us move faster. Demand in the critical minerals space has grown quickly in recent years, and we want our users to get fast, accurate answers.We're not starting from zero: our core system is already in active development, so you'll be building on real infrastructure from day one, working side by side with the team who designed and built it. You'll have room to help shape and extend it further, rather than sitting downstream of a larger team. Expect fast iteration and direct ownership, hands-on from the start. ResponsibilitiesFine-tune and adapt pretrained foundation models for domain-specific tasks on multi-dimensional imaging dataWork with the analytics team to translate their domain knowledge and requirements into concrete ML objectivesBuild robust training and inference pipelinesDesign experiments with measurable metrics, run ablation studies, and conduct failure analysisSupport work on RAG pipelines that inform model behavior as well as power analyst-facing queryingCollaborate directly on architecture and modeling decisionsMust-have qualifications3+ years of professional or equivalent hands-on experience building and training deep learning modelsProficiency in Python and PyTorchStrong fundamentals in deep learning and computer visionExperience training and fine-tuning pretrained or foundation modelsExperience with embedding models, including extraction and downstream useAbility to design experiments, define measurable metrics, run ablation studies, and conduct failure analysisExperience working with multi-dimensional image dataAbility to reproduce and extend research papersUnderstanding of transformer and attention-based vision architecturesExperience with embeddings, vector search, or retrieval systems that feed into model behavior. Project-level experience counts.Strong communication skills, with the ability to clearly document experiments, findings, and technical decisionsNice-to-have qualificationsDegree/coursework in computer science, machine learning, or a related quantitative field, or equivalent practical experiencePublications or contributions in deep learning, computer vision, geoscience, remote sensing or related venues (conferences, workshops, arXiv)Familiarity with structured scientific or sensor data in signal-heavy, high-dimensional domainsExperience with a vector database, such as FAISS, Pinecone, or QdrantExperience with self-supervised pretraining or frozen-backbone fine-tuningExposure to pretraining models from scratch, including data curation, self-supervised objectives, and training at scale, even at small or academic scaleAny other skills that could benefit a small, cross-functional team even outside deep learning, we'd love to hear about themRoom to growIf existing foundation models don't meet our needs, this role could extend into ground-up model pretraining, covering architecture design, self-supervised objectives, and training infrastructure. This isn't a day-one deliverable. It's a direction we're seriously evaluating, and candidates excited by that possibility will stand out.Why Esper?You'll be a foundational member of a fast-moving team, with direct impact on the quality of data we deliver to the world. We're a flat team, i.e., short paths from idea to decision, and your input shapes the product from the very first day. We value high-agency people who collaborate well and push themselves to build a better product.Ready to build the intelligence powering our next-generation hyperspectral systems?