The founders of ALMA, a LegalTech startup aiming to transform the immigration industry, were struggling to recruit a technical founder and team. ALMA's founders were deciding between us and another engineering firm for technical support. After choosing us because of our AI expertise, we stepped in to swiftly build their initial product and vet a suitable CTO in just three months.

‍

Product Vision and Specifications

ALMA's founders envisioned an AI drafting tool tailored to immigration lawyers. Understanding their needs, we identified the following key product specifications:

• Users can upload client context (PDFs, Word docs, links, etc.)
• Accurately retrieves necessary inputs from context
• Generates consistent and quality-checked documents
• Highlights potential hallucinations
• Provides an editable text interface
• Tracks customer edits
• Supports asylum and employment-based use cases

‍

‍

Development Journey

In building their MVP, we used React for the frontend and Next.js for the backend. We improved the AI engine quality by performing trial and error with different large language models (GPT, Claude, etc.), prompt chaining, the prompt itself, and different chunking strategies— including an alternative vector database and RAG approach. At one point, the project was delayed due to scope misalignment. Given that an urgent presentation was coming up with one of ALMA’s partners, our team worked over the holiday season to make up for the delay.

‍

Key Outcomes

‍

AI Legal Document Drafting Tool

Our final product displayed comprehensive context handling, document generation and consistency, customer interaction and editing, and support for two immigration-specific use cases.

‍

‍

Vetted and Suitable CTO

We supported choosing a CTO with the right balance of technical expertise and experience building large, enterprise software. Further, they exhibited an iterative, startup mindset.