About the role
About Eightpoint
Eightpoint is an internet technology company specializing in the agile development of products and content that address real-world interests, captivating users and driving significant growth for partners. With offices in the United States and Cayman Islands, Eightpoint collaborates with partners globally on the next generation of user-centric offerings. Our growing ecosystem includes innovative products like Weather Now, a sleek app that delivers real-time forecasts with clarity and ranks among the top three most-used weather apps in the U.S.; Check Heart Rate Now, a quick and easy wellness monitor; and Wave Browser, a powerful and secure way to search the web. Every product we launch is designed to engage users, enhance daily life, and deliver real-world value. Backed by data and driven by a relentless commitment to quality, Eightpoint moves fast, thinks big, and builds digital experiences that people love.
Our innovative culture thrives on collaboration, data-driven decisions, and a startup mindset. Join our team and enjoy top-tier benefits, unlimited PTO, flexible schedules, free lunches, and the opportunity to work alongside a world-class team of professionals.
About the Role
Eightpoint builds consumer products used by millions, on mobile and the web. We need an engineer who can turn our data into specialized models, from compact language models fine-tuned for specific tasks to predictive models built with our data team. You'll pick the right model for each problem and make it accurate, fast and cheap enough for production, whether it runs on a phone or behind a web service.
There’s also room to help build internal tools that use agentic AI to automate workflows.
You'll be involved across the whole lifecycle: framing the problem, building the dataset, training, evaluating, deploying and improving the model after it ships.
What you'll work on
- Fine-tuning compact transformer models for text generation and non-generative classification
- Distilling large models into small task-specific ones that fit size, latency and cost constraints - on device and on our servers
- Building training and evaluation datasets, including synthetic data, and keeping them clean as the label set grows
- Implementing deterministic guardrails to ensure accurate output
- Deploying models on device (LiteRT/TFLite, ONNX) and behind low-latency web APIs, shipped through staged rollouts
- Building MLOps pipelines that feed live production data back into retraining, evaluation and monitoring
- Building predictive and ranking models with our data team on large-scale user and product data
How you work
We’re looking for a self-directed engineer who thrives on ambiguous, open-ended problems, comfortable taking a loosely defined challenge, forming hypotheses, experimenting quickly, and driving it to a working solution with a high degree of autonomy.
Your input will shape product decisions beyond the model itself, including UI/UX and how features reach users.
Requirements
- 4+ years of professional ML experience in Python and ML Frameworks
- Hands-on fine-tuning of transformer models for language tasks, including parameter-efficient methods (ex: LoRA/PEFT)
- A working understanding of encoder-only and encoder-decoder architectures
- Knowledge distillation, or other ways of specializing a smaller model for one task
- Building datasets: sourcing, labeling, generating synthetic data and preventing train/test leakage
- A rigorous approach to model evaluation: meaningful metrics, calibration, and regression testing before every release
- SageMaker or similar cloud based training and deployment
Nice to have
- Edge deployment: quantization, LiteRT/TFLite, ONNX, Core ML
- Serving models in production web services at consumer scale
- Predictive modeling, recommendation or ranking on large behavioral datasets (BigQuery or similar)
- MLOps: containerized training, CI/CD for models, experiment tracking
- Public work on Hugging Face or GitHub
- RAG, embeddings and vector databases