Assit.io
A LLM-powered enterprise security solution, automating data representation via embeddings, eliminating manual normalization. Semantically rich embeddings drive efficient, minimally trained detections. The user-friendly natural language interface ensures accessibility. Gain benchmark access, connecting MSSPs and customers for data transformation from raw to enriched Alerts, supported by a robust architecture. Elevate your security strategy with our powerful solution
United States
2023
Domain
Security technology
Segment
B2B
Challenge
The primary challenge of this solution lies in effectively training and fine-tuning the Large Language Model (LLM) to create semantically rich embeddings that accurately represent security telemetry data. Achieving optimal performance in automated detection with minimal training examples while ensuring user-friendly natural language interaction poses a complex technical hurdle. Additionally, maintaining scalability and adaptability in the face of evolving security threats and requirements remains a continuous challenge
Results
User-Friendly Interface: Our UX design has created an intuitive, accessible interface that allows all users, regardless of expertise, to interact naturally with the system
Improved Experience: Our user-centered design greatly enhances the overall user experience by enabling human-like natural language interactions, reducing learning curves, and increasing user satisfaction
Efficient Workflows: We've streamlined complex workflows by integrating semantically rich embeddings into the UI, simplifying access to crucial data-driven insights for better decision-making.
Adaptable and Scalable: We've tackled adaptability and scalability challenges with a flexible mapping interface, ensuring system robustness and responsiveness to evolving user needs and security demands
Process:
Lightning Talks
On the kick off day we began the design process with lightning talks from the stakeholders involved to understand their perspectives on the problem being presented and understand their motivations and drive.
Understanding the domain
As I ventured into this entirely new domain, I initiated domain research to gain exposure. I delved into understanding essential terminologies and compiled a key takeaways document, enabling my fellow designers to reference and familiarize themselves with the domain
This allowed me not only to revisit the recordings but also to gain a deeper understanding of the problem statement
Primary & secondary research
Following the lightning talks, I extracted key insights and leveraged AI tools for my primary research. Additionally, I explored traditional products within the same domain to analyze their functionality and identify the primary challenges they currently encounter
This not only facilitated my comprehension of the current user pain points but also provided insights into the operation of these products and a deeper understanding of their core functionality
Golden Path
With a clear understanding of the pain points faced by the potential users and the main areas that need to be focused on, I worked with the stakeholders to start building a golden path for the MVP
This path represented the how the user will go through the product to reach the most valuable feature of the product
Sketching
After outlining the design path, I sought to create low-fidelity designs. To align with stakeholder expectations, I facilitated a 'crazy 8’s' sketching exercise, allowing them to visually communicate their vision and insights
This helped me not only understand what mental models they had in mind but also prepare solid reasoning beforehand if I was bringing designs that were radically different from what they were expecting.
High Fidelity Designs
With a well-defined clarity resulting from our process thus far, I embarked on crafting the high-fidelity version of the product
Creating a zero-to-one product doesn't mean aiming for the best version right away. Instead, we build upon initial assumptions during development, validate our ideas, and continuously iterate on the product, much like sculpting a work of art
