Tim DraperFounder and Managing Partne, Draper Associate · I back the ideas that look crazy first

Qazi AI is tackling one of the world’s most complex problems: access to justice.

2 Apr 2025 · 1 min

Watch on YouTube

1:07

In this brief panel discussion, the speakers outline Qazi AI’s approach to building a justice‑focused large language model. Speaker B explains that the company is customizing an existing LLM rather than using an off‑the‑shelf model, and highlights current engagements with the Government of Pakistan, prospective EU and U.S. State Department funding for a B2G deployment, and collaborations with law firms for B2B and B2C use cases. Speaker A probes the “special sauce” – the proprietary data and training methods that make the model uniquely valuable. Speaker C raises concerns about user adoption among non‑tech‑savvy villagers dealing with land disputes, prompting a discussion of accessibility and outreach. Finally, Speaker B notes that the model is currently trained in English but plans to add local language support, citing talks with a partner to incorporate Urdu. The conversation emphasizes a multi‑pronged go‑to‑market strategy that blends technical customization, strategic partnerships, funding pursuits, and localization to address access‑to‑justice challenges in developing regions.

What's the special sauce for how you train data, what proprietary data to make this really, really unique and not just something anybody can recreate?
So we are building on top of the model and not exactly using the same model. We are already engaged with the government of Pakistan. We are trying to secure funding from EU and State Department for B2G model. Similarly, a couple of law firms that we are already working in for B2B and B2C case are very much in.
When we are looking at consumers in small villages who go through these land disputes that you spoke about, They might not be even actually computer savvy. How are you thinking about the user adoption?
Bringing a very interesting perspective, like from the developing country background. She seemed interested, I think, but can't say much.
Right now we are training the LLM on English, but we plan to train it in on local languages as well. For example, we are already in talks with a company for Urdu adoption. We might not be able to solve their ultimate problems at the get-go, but it will be good information to start with.