Role: Lead Product Designer (Acting Cross-Functionally with ML & Tech)

Timeline: 3 Months (12 Weeks)

Platform: Web App (SaaS)

Prototype: https://www.figma.com/proto/gJSALrp1dGhe4L4YIZfbkc/Veritas-AI?page-id=2%3A48998&node-id=11-25722&viewport=-2875%2C-53%2C0.55&t=rJcnLlgpRZcKDizL-1&scaling=contain&content-scaling=fixed&starting-point-node-id=11%3A25722&show-proto-sidebar=1

https://www.figma.com/design/gJSALrp1dGhe4L4YIZfbkc/Veritas-AI?node-id=2-48998

The Business Problem & The ICP

Solo lawyers and small law firms in Malaysia operate with limited manpower. They lack the armies of chambering students that big corporate firms have.

• The "Before" State: Manually combing through CLJ Law or LexisNexis, cross-referencing Malaysian legislation, and organizing complex case files into chronological legal timelines would typically consume 30 to 50+ hours per complex case.
• The Goal: Build an AI-powered legal copilot (Veritas AI) that automates deep research and timeline generation without compromising absolute accuracy.

The Core Constraint: The AI Trust Deficit

Lawyers are inherently risk-averse. The biggest hurdle in legal tech is the LLM (Large Language Model) hallucination problem. If the AI hallucinates a nonexistent Malaysian caselaw, the lawyer could face severe professional consequences. Furthermore, heavy ML operations have high latency (slow load times) and high failure rates.

The UX & Product Solutions

To bridge the gap between AI capabilities and user trust, I designed three core pillars:

A. Grounded Chat with Verifiable Citations

I designed the primary chat interface to operate differently from standard generative AI. Every output is tied to a specific UI component displaying the exact Malaysian caselaw citation, allowing lawyers to instantly verify the source text without leaving the platform.

B. The "Independent Fact-Check" Audit Tool

The ML Challenge: AI models can sometimes be misled by their initial output, creating a self-reinforcing loop of bad data.