Login or sign up
For business
Average ratingGreat
4.1/ 5

Based on 9 reviews

544%
422%
333%
20%
10%
Ask AI

What reviewers say about Ulam.ai

Based on 9 reviews, listing data and website checks · updated 7 days ago

Customers describe the team as professional, knowledgeable, and cooperative. Reviewers frequently note that the company delivers projects on time and provides high-quality solutions.
What is Ulam.ai?

Ulam.ai is an AI consulting company based in Warsaw, Poland, that specializes in text-related machine learning solutions. They build data science strategies and offer one-on-one consulting services. Their work focuses on ultra-hard RL tasks and benchmarks in mathematics, STEM, and cybersecurity for frontier AI labs.

Who makes up the Ulam.ai team?

The company's team consists of PhD-level mathematicians and engineers who tackle research-level open problems. They create ultra-hard RL tasks and benchmarks designed to challenge leading models and help frontier AI labs train stronger reasoning capabilities.

What specific products or services does Ulam.ai provide?

Ulam.ai offers off-the-shelf datasets and custom problem sets for clients. They also build stateful environments that require tool use, multiple steps, and decisions based on intermediate results to help teams hillclimb popular benchmarks.

How does a new client get started with Ulam.ai?

Clients can start by hiring the company for a project, where they receive cooperative teams that deliver quality work. The process usually begins with providing sample RL tasks or exploring together which reasoning capabilities need to be strengthened before commissioning new tasks.

What benchmarks does Ulam.ai use for testing models?

The company provides ErdősBench, STEMBench, and CyberBench to test models on difficult problems beyond standard benchmarks. These benchmarks assess correctness, meaningful partial progress, and failure modes while building private evaluations around specific team needs.

What are the key features of their STEM and Mathematics RL tasks?

Their STEM RL tasks demand more than a plausible explanation, requiring quantitative predictions, computational results, and testable solutions checked against task-specific requirements. Mathematics RL tasks involve difficult problems with executable verifiers and precise reward signals where models must produce constructions or certificates.

How do their cybersecurity RL tasks work?

Cybersecurity RL tasks create challenging security environments for training and evaluating agents on investigation, defensive analysis, and tool use. These tasks include defined objectives and checks on task outcomes to measure meaningful progress in cybersecurity.

What is the current rating and review history for Ulam.ai?

The company has a 4.1 out of 5 rating from 9 reviews on Trustburn, with the most recent feedback dating back to November 2023. Reviewers frequently praise the team's high skill level, professionalism, and cooperative nature.

What do reviewers commonly praise about Ulam.ai?

Some reviewers on Trustburn mention that the team promptly addressed concerns and delivered projects on time, often exceeding expectations. Others noted the team was extremely helpful and supportive throughout the project lifecycle.

Where is Ulam.ai located and what markets do they serve?

The company is located in Warsaw within the Masovian Voivodeship in Poland, serving markets in the EU and US. Their website domain is ulam.ai and they maintain a LinkedIn profile for professional connections.

What is the core philosophy behind Ulam.ai's approach?

Ulam.ai states they focus on research-level problems that challenge leading models, distinguishing sound reasoning from plausible mistakes. They use expert review and task-specific checks to identify failure modes that further training can address.

How does Ulam.ai help teams improve their models?

The company helps teams find where models fail by evaluating mathematical reasoning, scientific research, and cybersecurity capabilities. They use these results to identify recurring errors and prioritize further training for specific capabilities.

How does the collaboration process work between Ulam.ai and clients?

Their collaboration process starts by providing sample RL tasks or exploring which reasoning capabilities need strengthening. From there, they can provide off-the-shelf datasets or build RL environments matched to those specific needs.

How long has Ulam.ai been in operation?

The website mentions they have been operating since 2017, as indicated by the copyright notice on their footer. They are an information services company focused on building data science strategies and 1-1 consulting.

How can I contact Ulam.ai?

You can reach Ulam.ai by e-mail at office@ulam.ai.

AI-generated from reviews on this page. Verify anything important with the company directly.

News

No news yet. This is where the business shares its own updates and press.

Photos

No photos yet. Photos shared by reviewers will show up here.

Widgets

Boost your business trust by displaying a widget on your website — a simple way to convert your visitors into customers.

Get the widget for your site →

About Ulam.ai

ulam.ai is an AI consulting company focused on text-related machine learning solutions. We specialise in building data science strategies and 1-1 consulting.

Website
ulam.ai