Everfit

Overview of job

About Everfit

Everfit is one of the fastest growing fitness coaching platforms in the world. We power more than 320,000 coaches across the globe with an all-in-one AI-powered platform to run and grow their businesses. In 6 years, we have become one of the leaders of the health and wellness B2B SaaS category in the US and around the world! We are at an inflection point. We are focusing our efforts to capture more of the $5.1 trillion health and fitness market by going direct to consumers. Our vision is to become the operating system for personal health, where anyone can plug in a coach, or multiple coaches, to reach their wellness goals faster, optimize their human potential, and extend their longevity.

Our team spans from San Francisco to Vietnam, with senior leaders who have built and scaled products at Netflix, Apple, Meta, Nike, Hulu, Uber, Amazon, Hypebeast, and Quibi. We are building the next unicorn to come out of Vietnam, and we are looking for a confident leader who can help us get there.

Position Overview:

You will train and adapt models using Everfit’s product data and feedback, then take them into production. You will own the technical work from problem definition and data preparation through training, evaluation, release, and monitoring.

Some problems call for a small classifier or ranking model. Others may benefit from adapting a pretrained language or vision model. You will assess where training can improve quality, personalization, speed, or cost, recommend an approach, and test it against the current product.

You will work with Applied AI Engineers, Backend Engineers, Data Engineers, and product partners. Your focus will be training data, learned representations, model adaptation, evaluation, and the model lifecycle. Applied AI colleagues contribute the surrounding context, tools, and application workflows. Together, you will make sure model improvements reach users and solve the intended problem.

Your impact:

Choose the problem and approach

• Work with product and domain partners to define the user problem, the decision a model should support, and how to measure a better result.
• Investigate whether errors come from the model, labels, missing context, reference data, or product integration.
• Compare the current workflow with simple baselines, existing models, and training-based approaches. Recommend the approach the evidence supports.

Build data that supports learning

• Work with data engineers and domain experts to prepare training and evaluation datasets. Define labels, review disagreements, and check missing records, duplicates, and uneven coverage.
• Turn product feedback into useful training examples. Check that a coach edit, click, purchase, or correction means what the training objective assumes.
• Keep training and evaluation separate. Use time, user, coach, or workspace boundaries where needed, and prevent future information from entering earlier examples.
• Test unfamiliar users and items, sparse histories, and important failure cases. Distinguish missing activity records from inactivity when working with engagement data.

Train, adapt, and evaluate models

• Build task-specific models and adapt pretrained models where appropriate. Methods may include supervised fine-tuning (SFT), Direct Preference Optimization (DPO), distillation, and parameter-efficient fine-tuning (PEFT), including low-rank adaptation (LoRA).
• Choose metrics that match the task. Examine ranking quality, per-category classification errors, language quality, or recognition and portion errors as appropriate.
• Check uncertainty and define when a model should abstain, request more information, or use a fallback. Review the consequences of different mistakes with product partners.
• Version datasets, labels, transformations, model artifacts, and evaluation settings so another engineer can reproduce the result.

Deliver and improve the product

• Work with backend, data, and DevOps colleagues to deploy and serve models in batch or online services. Keep training and serving transformations consistent and test the full path to the result users receive.
• Compare model quality with response time, memory use, and operating cost. Diagnose failures and make improvements without hiding regressions in an average score.
• Run controlled product experiments with the team. Measure useful outcomes alongside offline scores, and document what the evidence supports.
• Monitor model and data performance after release. Validate updates before deployment, maintain a rollback path, and turn confirmed failures into regression tests.
• Protect coach and client data in datasets, inference, and logs. Follow workspace access boundaries and approved retention and deletion rules.

Compensation & Benefits:

• Full social insurance coverage and premium healthcare benefits.
• Competitive salary with 13th-month pay and performance bonus.
• Paid leave — vacation, sick leave, and maternity leave

Build Your Career with Impact:

• Be part of a mission-driven company that’s transforming human potential.
• Empower over 300,000+ coaches in more than 200 countries - including Olympic athletes, military trainers, rehab specialists, and celebrity coaches.
• Make a global impact while building a meaningful, purpose-driven career.
• Continuous training, mentoring.
• 1-2 performance reviews annually.

Work Environment:

• English-speaking, international environment promoting open communication and collaboration.
• Work alongside world-class talent from companies like Apple, Meta, Uber, Patreon, and Nike, united by a shared passion for impact and excellence.
• International team for global career growth.
• State-of-the-art equipment provided.
• Beautiful working environment near Danang's city center (Dragon Bridge) and Hanoi office. 

Health & Wellness:

• PVI health insurance program and annual health check-ups.
• Active sports clubs (football, badminton, pickleball) and company retreats.
• A modern pantry designed for relaxation and social connection.
• Monthly wellness and learning workshops to recharge and grow.