Computer Vision / Machine Learning (CV/ML) Engineer Save

Work Location: Remote (U.S., excluding AK and HI)

Status: Exempt / 1.0 FTE (40 hours/week)

Reports to: Head of Product & Engineering

We are looking for a Computer Vision / Machine Learning (CV/ML) Engineer to help us improve the machine learning systems that power iNaturalist's species identification (vision API) and geographic range modeling (geomodel). Working alongside our senior backend engineers, you will contribute to the full ML lifecycle—innovation and experimentation, model training, evaluation, and production deployment—building, maintaining, and modernizing systems that help millions of people identify what they are seeing in nature.

The ideal candidate brings hands-on experience shipping CV/ML models in production (i.e., not just small-scale experiments), comfort working at the intersection of research and engineering, and enough ecological curiosity to care about getting the science right. You will have real ownership over our ML stack and will help us evaluate when and how to adopt newer model architectures and software tooling.

This role reports to the Head of Product and Engineering and works closely with other members of the Engineering team.

About iNaturalist

iNaturalist is one of the world’s largest biodiversity platforms—powered by millions of community scientists and supported by expert naturalists, researchers, and conservation organizations. With hundreds of millions of observations and 400,000+ active monthly contributors, iNaturalist data informs conservation decisions, accelerates biodiversity research, and inspires people everywhere to connect with nature.

Founded in 2008 and operating as a nonprofit committed to open data, iNaturalist blends crowdsourcing, machine learning, and community science to help people understand, document, and protect biodiversity. Our vision is a world where everyone can discover, understand, and help sustain life on Earth.

We are growing—and we are looking for people energized by purpose-driven work to help scale the platform and experience for millions around the world.

Responsibilities

• Own and operate our monthly CV model training cycle, producing updated species classification models across tens of thousands of taxa and publishing accompanying release notes for the iNaturalist community (e.g., blog posts documenting newly supported species and model performance).

• Maintain and improve our model training and serving infrastructure, working across the inatVisionTraining, inatVisionAPI, and inatGeoModelTraining codebases (currently Python / TensorFlow / Keras / Flask).

• Manage the full model lifecycle: data prep, training runs, evaluation, export, and deployment to web and mobile environments.

• Implement and monitor model performance in production, tracking accuracy and coverage regressions across taxa, geographic regions, and observation types.

• Manage and advise on hardware and cloud infrastructure, working in our current Linux hardware and advising on hardware improvements or cloud alternatives.

• Train and optimize on-device CV and geo models for iOS and Android, ensuring high-quality identification within strict size and latency budgets on our React Native mobile app.

• Apply quantization, pruning, and other optimization techniques to maintain accuracy while hitting mobile performance targets.

• Collaborate with the iNaturalist Mobile team to integrate and test updated models in the app.

• Lead development of Sound ID: design and implement a pipeline to support species identification from sound recordings.

• Work with the Ops team to deploy a large-scale “vision language model” system built by academic partners that enables text-based image retrieval.

• Build bounding-box detection capabilities to locate organisms within images, supporting improved cropping for downstream classification.

• Develop automated image quality assessment, such as detecting AI-generated images, flagging blurry images or observations lacking a clear focal organism, to help curate training data and surface feedback to users.

• Evaluate and integrate relevant advances from the research literature (e.g., vision-language models, improved backbone architectures) when they offer practical production benefit.

• Work with other engineers to maintain and improve training data pipelines, including ingesting and processing large datasets and handling class imbalance across taxa and geographies.

• Collaborate with our Engagement team and iNaturalist’s network of expert identifiers and researchers to validate model outputs and incorporate taxonomic updates.

• Contribute to the team’s engineering culture—using standard version control and ticket tracking (Github and Linear), writing clean, documented, testable code, participating in code review, and sharing knowledge across a small, fully remote team.

Qualifications

• 5-7+ years of professional engineering experience, including 3+ years of experience in machine learning engineering, computer vision, or a closely related applied ML role.

• Strong Python skills and hands-on experience with tools like PyTorch or TensorFlow/Keras (our current stack). 

• Demonstrated experience training and deploying CV models in production—at scale, with real users—not solely in academic or research settings.

• Experience exporting and optimizing models for mobile deployment (e.g., TFLite and/or CoreML), including quantization and performance benchmarking.

• Solid understanding of standard CV techniques: image classification, object detection, and image quality assessment.

• Comfort working with large, noisy, user-generated image datasets, including handling class imbalance, label noise, and geographic distribution shifts.

• Experience with audio ML or spectrogram-based classification.

• Experience with experiment tracking tools (WandB, MLflow, or similar) and disciplined approaches to reproducibility.

• Strong written and verbal communication skills—able to write clearly for both technical teammates and a non-technical community audience (e.g., model release blog posts).

• Comfort working autonomously in a small, fully remote team with significant ownership over your area.

• Familiarity with geographic/spatial ML (using, e.g., deep learning-based species range modeling).

• Background in ecology, biology, or natural history—or just a deep personal interest in the natural world.

• Experience building and operating vector search systems in production — familiarity with Qdrant or comparable vector databases (Pinecone, Weaviate).

• Experience generating and managing embeddings at scale, including multimodal embeddings (e.g., CLIP or SigLIP) and chunking/indexing strategies.

• Familiarity with distributed ingestion pipelines using Ray, Spark, or equivalent frameworks for parallel embedding and upsert workflows.

• Experience working with LLM APIs (e.g., Anthropic Claude, Google Gemini, OpenAi, self-hosted models) to analyze, summarize, or extract structured information from large text datasets.

Why Join Us?

• Impact: Play a pivotal role in a mission-driven organization dedicated to biodiversity conservation and community engagement.

• Innovation: Working with a team of developers building cutting-edge technology that empowers millions of people to make a positive impact on nature.

• Growth: Be part of a growing organization with opportunities for personal and professional development.

If you are passionate about biodiversity, technology, and working with other talented engineers to achieve meaningful impact, we encourage you to apply for this exciting opportunity.

Additional Information

A background investigation is required prior to employment for applicants who receive a conditional offer of employment. Applicants given a conditional offer of employment will be required to sign authorization and release forms enabling such an investigation. 

How to Apply & Timeline

Instead of a traditional cover letter, you will answer a short set of questions designed to assess your experience and problem-solving approach. Text inputs are limited to 250 words each. We want to see your experience and understanding. We don’t consider LLM outputs to be a substitute for your own experience and knowledge.

Responses are initially reviewed (by real people, not AI) without names and independently in order to reduce bias. Demographic information (optional) is used only in aggregate reporting.

The application will close on Jul 15, 2026. However, we may close it after receiving 150 applications. In this case, we will notify anyone who has started the application that they have 48 hours to complete it.

• Initial Application Review: Jul 15, 2026-Jul 22, 2026

• Written Follow-Up: Jul 23, 2026

• Interviews: Jul 27, 2026–Aug 7, 2026

• Offer: Aug 12, 2026

• Ideal Start Date: Aug 24, 2026

If you're excited to help build tools that help millions of people engage with nature, we’d love to hear from you.