The Opportunity
At Dyrt, we’re building the future of food waste intelligence. From our in-vessel aerobic composting facility just south of Downtown LA, we’re developing platforms that help large businesses understand, reduce, and track their food waste. One of those tools is Dyrty Vision, our machine learning platform that generates qualitative insights from real-time images taken above our sorting line.
We're looking for a Vision Machine Learning Intern who can independently drive the development of a commercially viable vision-based prototype, from data collection to cloud deployment. You’ll work at the intersection of sustainability, AI, and hardware, taking ownership of projects that ship quickly and have immediate environmental impact.
What You'll Do
- Own a full-stack vision ML project: propose, prototype, and deploy a working system using live data from our composting facility.
- Build and train models to classify, cluster, or segment waste types and contaminants using modern vision techniques.
- Engineer the full pipeline: from image acquisition and preprocessing to inference and data presentation in a customer dashboard.
- Collaborate cross-functionally: join ops team lunch to understand real-world friction, then turn that insight into product improvements.
- Work closely with our head of product and lead developer to turn models into usable tools for our customers and warehouse staff.
Tech You'll Use
Computer vision frameworks (e.g. PyTorch, YOLOv8, Roboflow, OpenCV, Segment Anything)
Cloud deployment tools (e.g. FastAPI, Supabase, Vercel, GCP or AWS)
Data collection tools for video-to-frame pipelines and labeling interfaces
Annotation platforms (e.g. Label Studio, CVAT, Supervisely)
Experiment tracking (e.g. Weights & Biases, TensorBoard, Comet.ml)
Dataset management and analysis (e.g. FiftyOne, Hugging Face Datasets, Albumentations)
Model export and inference (e.g. ONNX, TorchServe, NVIDIA Triton)
Edge deployment (e.g. NVIDIA Jetson, Intel OpenVINO, Coral TPU)
Task queue and streaming infrastructure (e.g. Redis, Celery, Kafka, RabbitMQ)
Visualization and dashboarding (e.g. Streamlit, Dash, Plotly, Chart.js)
You Might Be a Fit If You...
- Have built your own ML project end-to-end - whether that’s for a class, a hackathon, or just for fun.
- Have experience training, tuning, and evaluating image classification or segmentation models.
- Are curious about deploying vision systems in messy, real-world environments (not just clean academic datasets).
- Can identify bad UX, understand user pain points, and care about solving them, not just coding around them.
- Love figuring things out and don’t mind jumping between dev, research, and product hats.
- Bonus: you’ve explored combining vision + LLM pipelines to enhance labeling or explainability.
What You’ll Gain
- A portfolio-worthy vision ML prototype solving a real business problem
- Experience deploying real-world ML systems
- A crash course in product thinking and startup execution
- Firsthand experience in sustainability tech and waste infrastructure
- Ongoing mentorship from experienced product and engineering leads
- The chance to pitch your project to our investors at Demo Day
Logistics
- Location: On-site at our warehouse near Downtown LA
- Schedule: 3 days/week on-site (flexible)
- Dates: June 2–9 start | Aug 8–15 end
- Pay: $20–$30/hour depending on experience
- Perks: Weekly team lunches, sustainability field trips, partner event invites, potential for extended work
How to Apply
No Formal Cover Letter Required. Just Tell Us
- Who you are and what you’re passionate about
- One project you’ve worked on that you’re proud of — and why it matters to you
- Links to your portfolio, GitHub, or anything else that showcases your work
- In 2–3 sentences, pitch an idea you might explore with Dyrty Vision this summer
- (Optional) Tell us one thing you’ve upcycled Deadline: May 5, 2025
Interviews: Rolling basis — apply early!
