‘Next billion-dollar company’: Aporia stands as a leader in AI operations, ensuring your AI journey is marked with trust, visibility, and scalable growth

“As the market becomes saturated with new AI applications, the company is providing reliable infrastructure and tools for observability, monitoring and optimizing AI, ML and Large Language Models (LLMs)”

Aporia is the AI/ML observability platform that empowers data scientists with essential tools to monitor, manage,  investigate, and improve machine learning models in production. The platform seamlessly supports every ML use case and all model types, enabling ML teams to tailor Aporia to their specific needs. It offers deep model visibility and comprehensive control in detecting issues such as data drift, bias, data integrity problems, and performance degradation. By deriving powerful insights, Aporia ensures trustworthy and responsible AI that benefits businesses and society. Notably, Aporia has earned the trust of Fortune 500 companies and data science teams, including B/S/H, Lemonade, Munich RE, SIxt, and Armis, spanning various industries worldwide. 

The Silicon Review reached out to Liran Hason, co-founder and CEO of Aporia, and here’s what he had to say.

Interview Highlights

Q. What inspired the creation of Aporia?

As a former Machine Learning Architect at Adallom (acquired by Microsoft), I directly experienced the significant risks associated with deploying ML models to production without adequate guardrails. As an investor for Vertex Ventures, I also guided organizations through the challenges of managing, maintaining, and ensuring the value-added by these models. These experiences were instrumental in shaping Aporia.

Now, as the CEO of Aporia, I apply my first-hand knowledge to our work every day. We built Aporia from the ground up to provide safety and certainty to the complex process of monitoring and managing ML models in production.

Q. As the ML Observability platform, what are Aporia’s key focus areas?

Aporia plays a pivotal role in enabling organizations to trust and scale their AI across various fields such as pricing, fraud detection, marketing, supply chain, retail, human resources, credit risk, insurance, and more. The company focuses on the following three key areas:

Reliability: Our primary objective is to ensure consistent peak performance ML models in production, driving revenue growth. Aporia centralizes model management and offers comprehensive monitoring to detect model drift, degradation, staleness, bias, and data integrity at scale. This enables early identification of potential issues, allowing you to take corrective action before they cause a significant impact, saving time, resources, and maintaining user trust.

Transparency: Aporia is on a mission to infuse AI with trust. Our platform enables you to delve into the intricacies of your models in production and explains their predictions. By providing a high degree of visibility, Aporia empowers you to rationalize model decisions, thereby fostering trust among stakeholders. Furthermore, Aporia supports your AI governance efforts by offering clear and understandable model documentation and decision-making process, promoting ethical and responsible AI usage.

Scalability: Aporia is designed to scale seamlessly with your needs as the complexity and number of ML models increase, without compromising on user experience. It serves as a central hub for all your models, ensuring easy management and visibility regardless of the scale. With Aporia’s easy setup and quick integration, you seamlessly move from minimal production insight to robust ML observability.

Q. Can you introduce us to your services? What are their primary features?

Our primary offering is the continuous monitoring and maintaining of machine learning models in production. Aporia

allows for rapid integration, enabling model monitoring and visibility within minutes, and allowing direct connections to various data sources to ensure data integrity.

Aporia’s ML observability platform consists of the following four main pillars:

Model Monitoring — Detect and alert ML teams of drift, bias, degradation, staleness, performance, and data integrity issues.

Visibility — Centralize model management under a single hub and track the behavior and performance of your models in dashboards tailored to what matters to you.

Explainability — Ensure model transparency, understand feature impact, and simulate ‘what if’ scenarios to gain insights into t how your model reaches specific predictions.

Root Cause Analysis — Analyze production output to uncover new connections in your data, or quickly identify the root cause of any production issues.

All these features together provide a comprehensive ML observability solution for managing, monitoring, and improving ML models in production environments.

Q. Can you provide us with one or two success stories describing the challenges your clients faced and how Aporia helped them overcome those challenges?

Background: Our customer is a leading eCommerce platform that caters to a wide range of fashion-forward customers. They utilize machine learning for two primary functions: Product recommender systems and an LLM feature where users can textually ask the system to organize a wardrobe.

Challenges: The customer experienced several challenges with their existing ML infrastructure. False positives were a significant issue, leading to misidentified product recommendations and wardrobe suggestions. This not only frustrated customers but also led to a decline in sales conversions. Additionally, the customer was grappling with time-consuming manual monitoring and management of their ML models, hampering their ability to respond to performance issues. They often only discovered drift weeks after it occurred, resulting in revenue loss.

Aporia Integration: The customer implemented Aporia's ML Observability platform to mitigate these challenges. Aporia's quick and seamless integration process, leveraging Direct Data Connectors (DDC), allowed the customer to connect to their existing data sources without the need for risking data duplication or changes to production code. The entire integration process took minutes, saving the customer significant time compared to legacy solutions.

Results: After implementing Aporia, the eCommerce platform saw substantial improvements:

• False Positives Reduction: By leveraging Aporia’s model monitoring, the customer promptly detected and addressed model drift and bias issues. Aporia's production alerts flagged potential issues that could lead to false positives, allowing the team to take corrective action immediately. This led to a 30% reduction in false positives within the first three months of implementation.
• Time Savings: Using Aporia's custom dashboards, monitors, and investigation tools significantly reduced the time spent on manual model monitoring and investigation. The team estimated saving approximately 20 hours per week, allowing data scientists and ML engineers to focus more on strategic tasks.
• Performance Enhancement: Aporia's platform enabled early detection of performance degradation, leading to faster resolution times. As a result, the customer saw a 15% improvement in the accuracy of their recommender systems, leading to increased sales conversions.
• Revenue Increase: With improved model performance and reduced false positives, the eCommerce platform saw a surge in customer satisfaction and sales conversions. Within the first six months post-Aporia implementation, the company reported a 20% increase in revenue from the recommender system.

Q. What new endeavors is Aporia currently undertaking?

Aporia is at the forefront of pioneering innovative and collaborative methods for centralized ML model visibility, monitoring, and root cause analysis. It enables teams of all sizes to easily pinpoint the origin, timing, and reasons behind production issues, Moreover , Aporia facilitates segment and drift analysis to uncover new connections in production data.

With the increasing use of generative models, like chatGPT, Aporia is also dedicated to providing advanced troubleshooting capabilities for unstructured model types. These capabilities include embedding visualizations and prompt clustering, which enable users to assess performance, detect bias, and effectively mitigate hallucinations in large language models.

Q. Is there anything you would like to add before we wrap up?

AI has the potential to be the best thing that’s happened to humanity or the worst thing. However, with the proper guardrails and responsible AI practices in place, businesses, users, governments, and society at large will be able to enjoy the immense benefits AI systems offer to improve our lives.

Leadership | Aporia

Liran Hason, Co-founder & CEO: Prior to founding Aporia, Liran was an ML Architect at Adallom and later an investor at Vertex Ventures. He created Aporia after seeing first-hand the effects of AI without guardrails. In 2022, Forbes named Aporia as the ‘Next Billion-Dollar Company’.

Alon Gubkin, Co-founder & CTO: Alon, has led Aporia in raising $30 million from investors like Tiger Global Management and Samsung Next. In 2022 and 2023, Alon was named to ‘Forbes 30 Under 30’.

“Our principal objective is to ensure your ML models in production deliver peak performance consistently, thereby driving revenue growth.”