Machine Learning on a Typewriter

Perhaps you've heard about our AI Search, launched last November: a powerful new way for clients to quickly find the perfect freelancer for any project, even easier and better than ever before. AI Search no longer relies solely on keywords, but on the entirety of your profile to make a perfect match. 

What you may not know, however, is that teams at Malt have been working diligently on our small language model for a while now. Malt's AI capabilities were built in-house, and are not simply powered by ChatGPT or another large language model (LLM). LLMs have broad applications that can have specific features added to them to handle certain tasks. Unlike general-purpose AI, we’ve leveraged 11 years of internal data and millions of matches to develop proprietary algorithms specifically tailored to the nuances and complexities that come with hiring the right people at a scale. Malt's own Machine Learning and AI researchers (Warren Jouanneau, Marc Palyart, and Emma Jouffroy) presented their findings last October at the ACM conference on Recommender Systems via a research paper, entitled “Skill matching at scale: freelancer-project alignment for efficient multilingual candidate retrieval.” Let's unpack their findings in order to better understand the advancements that have been made at Malt in terms of AI. 

The Challenge: Matching at Scale

Malt connects more than 800,000 freelancers with businesses seeking their skills and expertise. With so many professionals and projects on the platform, ensuring the right match is a complex challenge—especially when working across multiple languages, countries and varying projects.

Over time, we identified a few aspects that we were looking to improve:

  • Our system struggled to scale efficiently as new projects were constantly added.
  • It didn't fully utilize the rich details in freelancer profiles, sometimes overlooking key skills and experience.
  • It was language-specific, meaning a separate matching system had to be built and maintained for each language.

We wanted to make it easier and faster for hiring managers to find the right freelancer for any project, so we decided to introduce AI-powered skill matching into Malt. 

Powered by AI: Smarter, Faster, and More Precise

To improve the matching process, Malt’s research team developed a sophisticated AI model that analyzes both freelancer profiles and project descriptions using multilingual machine learning. We built what is called a "neural retriever” in Malt to scale freelancer matching, which is essentially a way to organize our data so that freelancers with similar skill sets are digitally stored close to each other. The end result is a mapping of our data, which you can see a visualization of below. By doing this, we are able to find the right freelancer candidates very quickly once we can identify where a project belongs in the map.

Mapping freelancer data

The new system works in three key ways:

  1. Understanding Context Better: Instead of just looking at keywords, the AI interprets the full meaning of project descriptions and freelancer profiles. This means it understands nuances like industry jargon and synonyms.
  2. Multilingual Matching: The AI can process job descriptions and profiles in multiple languages, making it easier to connect freelancers and businesses across different countries.
  3. Learning from Past Successes: The system improves over time by analyzing which past matches led to successful collaborations. If similar projects have been successfully completed by freelancers with certain skills, the AI can recommend similar profiles for future projects.

How Does This Help You as a Freelancer?

This new system isn’t just an improvement for Malt—it directly benefits our whole community by providing more relevant project matches and improving your visibility. The AI analyzes a broader range of your profile information to ensure you’re matched with projects that truly align with your skills and experience, reducing irrelevant opportunities. It also enhances your profile’s visibility, surfacing it more effectively for potential clients and increasing your chances of being discovered. You can rest assured, however, that as Malt's AI was designed internally, it was developed with our freelancers in mind.  Our commitment to our community is inscribed in our AI guiding principles, which include using AI responsibly as a way to assist humans and not replace them. We also commit to using AI in order to make transparent recommendations to hiring managers, and to always preserve the power of choice for both clients and freelancers alike. 

How to Optimize Your Profile for AI Search

Now that you know how the new AI matching system works, here are some practical steps you can take to improve your chances of getting the best project recommendations:

  1. Fully Complete Your Profile: The system uses all available information to make accurate matches, so ensure your profile is as detailed and up to date as possible, including all of your languages, training and certificates, as well as your geographical availability. 
  2. Use Clear and Specific Job Titles: Instead of generic terms like “Consultant,” use specific ones like “Digital Marketing Consultant” or “JavaScript Developer.” This will help the AI to best identify your profile and map you in the correct digital space for your ideal project opportunities. It will also make it easier for clients to differentiate you from other freelancers with similar profiles.
  3. Describe Your Experience in Detail: Instead of simply listing past job titles, describe the key responsibilities and skills you used. In addition, be sure to regularly update this section of your profile with your most recent experiences. 
  4. Keep Your Skills List Updated: Skills are trending everywhere in 2025: update your listed skills to reflect your latest expertise and include both broad and niche skills relevant to your work. Be sure to offer some context, whenever possible, on how you learned or utilized your skills by showing the specific project you worked on.
  5. Respond to Project Offers Promptly: The AI learns from freelancer interactions, so engaging actively with project offers can improve your ranking over time. Our data shows that 70% of clients start shortlisting candidates within 24 hours. 

A Future of Even Smarter Matching

Since launching AI Search last November, we've seen a rapidfire increase in the time it takes clients to find the right freelancer for their project: the average length of time has been divided by three, and it now takes less than a day to hire a freelancer. 

But this is just the beginning. Future improvements may include even more accurate personalization, integrating freelancer preferences, and refining recommendations based on long-term career growth. 

At Malt, we’re committed to making freelancing easier, more efficient, and more rewarding for everyone. Our dedication translates to continuously pushing the envelope by reinvesting in our product, paving the way to allow technology to make hiring and sourcing a seamless process for both clients and freelancers. 

With this AI-powered system, you can spend less time searching for the right projects and more time doing what you do best—delivering great work!

Update your profile today!

Ready to make the most of the new matching system? Update your profile today and see the difference!