Job Specifications
Who we are
We help consumers shop smarter. Our ambition is to revolutionize the way people shop, empowering them to find what they need, make informed purchasing decisions, and save money.
Joko is a tech company founded in Paris. Our team of over 80 talents is international and spread across offices in Paris, Barcelona and New York, or working remotely.
As a certified B Corporation, we are committed to making a difference in the lives of the communities we serve, including the 4 million consumers already using Joko to save money every day at over 7,000 merchants.
Over the years, we have quickly expanded our value proposition to enable consumers to make smarter shopping decisions through many features: cash back, buy now and pay later, automatic coupons, price drop alerts, carbon footprint tracking, and more. Today, we are dedicating significant resources to developing an AI-powered assistant that helps users find the right product based on their criteria for price, quality, or environmental impact from the global e-commerce catalog.
After reaching profitability in our core market, we are now expanding internationally with a primary focus on the US.
It's still day 1; come build the future of shopping with us!
This position will be located in Paris
Research at Joko
At Joko, we believe that today's online shopping experience is fundamentally flawed. Our ambition is to disrupt it by enabling users to find their desired products in the smoothest way possible, to effortlessly compare all their characteristics, and to access transparent and clear information on both their price and environmental cost.
To achieve our goal, our research team tackles two main challenges:
Building the universal product catalog : We are developing the world's largest universal product catalog, bringing together all the products sold across e-commerce worldwide. This requires the ability to understand any web page in order to extract structured information, and to clean and standardize data from millions of heterogeneous sources in near real time. One of the biggest difficulties lies in scale: we must process hundreds of millions of products multiple times per day
Developing the AI Shopping copilot: On top of this catalog, we are building an AI copilot that helps users navigate our product universe through natural conversation. This goes well beyond traditional RAG systems: it requires combining full-text and vector search across massive datasets, while leveraging LLMs to ensure an experience that is reliable, fast, and cost-efficient. We continuously iterate to improve search accuracy, reduce latency, and better capture user intent.
To meet these challenges, we rely on LLM-based methods. We have developed state-of-the-art approaches centered on fine-tuning relatively small models, striking a balance between performance and efficiency. Still, much research remains to further optimize resource usage, scale our systems to ever larger volumes of data, and explore new architectures that push the boundaries of what LLMs can achieve in large-scale e-commerce applications.
What you'll do
At Joko, our Machine Learning interns are considered full members of our engineering teams. We design internships to be deeply tied to real-world projects, ensuring that interns can have a tangible impact from day one.
For end-of-studies internships, our goal is not only to provide a meaningful learning experience but also to prepare interns for a smooth transition into full-time engineering roles at Joko after their internship.
You will have full ownership of your project and the freedom to orient your research direction based on your results and what you consider promising. Your ultimate goal will be to deploy your work in production and monitor its impact on hundreds of thousands of users. Exploration is at the heart of the internship: you will conduct experiments, literature reviews, and theoretical developments.
As a Machine Learning Research Intern, You Will Work With Our Engineering Teams To Maximize Your Impact, And You Will Work On One Of The Following Research Topics
Improving the Universal Catalog Pipeline
Enhance the performance and scalability of our LLM-based data processing pipeline
Explore fine-tuning for specialized tasks and techniques to reduce model size (quantization, pruning, distillation)
Develop rigorous evaluation methods, notably using LLMs as judges
Improving Search in the Catalog
Benchmark search techniques that combine full-text and vector search
Explore using images to enhance search performance
Identify and evaluate the most effective LLM-based embedding methods
Improving the AI Copilot
Benchmark and fine-tune models to improve both performance and latency
Work on model optimization (size reduction, distillation, quantization)
Establish rigorous evaluation pipelines
Your Responsibilities
Research: Formalize objectives, conduct literature reviews, design new algorithms, and collect/transform data for experim
About the Company
We help consumers shop smarter. Our ambition is to revolutionize the way people shop, empowering them to find what they need, make informed purchasing decisions, and save money.
Joko is a tech company founded in Paris. Our team is international and spread across offices in Paris, Barcelona and New York, or working remotely.
As a certified B Corporation, we are committed to making a difference in the lives of the communities we serve, including the 4 million consumers already using Joko to save money every day at over 7,000...
Know more