Concept

Jigsaw is a SAAS tool for product design teams that accelerates the wireframing process by generating context-aware design drafts. Unlike other generative UI tools that produce generic templates, Jigsaw integrates directly with a team's existing Figma design system and their user research repository. By understanding a team's unique components and user pain points, it transforms a simple text prompt (like a user story) into multiple, relevant wireframe options that are already on-brand and data-informed.

Value Proposition

Stop starting from a blank canvas. Go from user story to data-driven, on-brand wireframes in seconds, not hours. Jigsaw bridges the critical gap between research and design, ensuring every idea starts with your user and your style guide in mind.

AI Integration

Jigsaw is built on a Retrieval-Augmented Generation (RAG) architecture:

• Retrieval: When a designer enters a prompt, Jigsaw uses vector search to query two sources: the team's Figma component library (to find relevant UI components) and their indexed user research data—interviews, feedback, and notes (to find relevant user insights).
• Augmentation: The retrieved components and research insights are passed to a Large Language Model (LLM) as rich context.
• Generation: The LLM uses this context to generate a logical layout and structure for a wireframe, delivered as an editable draft directly within Figma. Each major design choice can be annotated with the specific user insight that informed it.

First MVP

The first Minimum Viable Product will be a Figma plugin with the following core features:

• A simple text area where a user can paste a user story or design brief.
• A second text area to paste relevant user research notes or key insights.
• A one-time connection to the team's active Figma file to read their component library.
• A "Generate" button that produces a new, editable lo-fi wireframe page in Figma using the correct components from their library.

Unique Selling Points

• Context-Aware Generation: Generates designs using your components and your user research, not generic templates.
• Traceability: Creates a clear, auditable link from a design element back to the user insight that inspired it.
• Workflow Acceleration: Drastically reduces time spent on initial ideation, wireframing, and ensuring design system consistency.
• Built for Teams: It doesn't replace designers; it empowers them by handling the initial, repetitive groundwork, allowing them to focus on high-level problem-solving and creative refinement.

Concept

LogLens is a SAAS tool designed for Autonomous Vehicle (AV) and robotics engineers that uses AI to dramatically accelerate the root cause analysis of system failures (disengagements). When an AV encounters a problem and disengages, engineers upload the event's complex log file to LogLens. The platform automatically analyzes the multi-modal data (video, lidar, sensor readings, system states) and generates a concise, human-readable summary of what went wrong and why.

Value Proposition

Reduce the time it takes to debug a single AV disengagement from hours or days to just minutes. We turn complex log files into interactive, understandable incident reports.

AI Integration

• AI Summary Generation: A multi-modal AI model processes and correlates data from different streams to identify anomalies and causal links. An LLM then synthesizes these findings into a plain-language summary (e.g., "The vehicle disengaged due to a 70% drop in the lead vehicle's perception confidence score immediately following a lane change on a wet road.").
• Conversational Q&A: After the initial summary, engineers can use a natural language chat interface to "interrogate" the log file. They can ask questions like, "Show me the LIDAR point cloud at timestamp 2:31.5" or "What was the planner's intended trajectory 3 seconds before the disengagement?" This replaces the need for complex, manual queries and data plotting.

Unique Selling Points

• Speed: Provides an initial root cause hypothesis in under 5 minutes, compared to the hours of manual analysis typically required.
• Accessibility: Empowers junior engineers to perform sophisticated diagnostics that previously required senior-level expertise.
• Focused Workflow: Purpose-built for the AV disengagement debugging loop, unlike generic log analysis platforms.

First MVP Product

The MVP will be a web application with the following core functionality:

• A user can upload a single, standard-format AV log file (e.g., a ROS bag file).
• The backend processes the file, running AI models to detect key anomalies and generate a summary.
• The output is a single, non-interactive report page that includes:
  • The AI-generated event summary.
  • A timeline view highlighting key anomalous events.
  • Embedded video clips and data visualizations (graphs) corresponding to the highlighted events.

This MVP validates the core value proposition of automated analysis and provides a foundation to build the interactive, conversational features in the next iteration. It is a tightly scoped product that a small team of engineers can build and launch quickly.