§ Products · DMA-PRD-001

We build AI products end-to-end
for ourselves and our clients.

The two below are ours. They run in production right now. They are the proof of what the team can do.

Beta · public
LegalPro

AI legal research for Indian courts. 6M+ High Court judgments indexed and searchable in plain English. Semantic search, automatic summaries, citation graphs, predictive analysis — for solo practitioners and small firms.

Why it exists: Indian lawyers still rely on Manupatra and SCC — expensive, keyword-based, and slow. LegalPro indexes the same corpus with AI-grade semantic search and surfaces case law the way a sharp junior lawyer would, not an old database would.

LegalPro screenshot
LegalPro · features

Semantic search

Ask in plain English. Not keywords. The engine understands intent, finds related cases even when the exact phrase doesn't appear.

AI summaries

Every judgment summarised automatically — facts, ruling, reasoning. Built on models fine-tuned on Indian legal text, not Western legal corpora.

Citation graphs

See which cases a judgment relies on, and which later judgments cite it. Built across decades of HC decisions.

Predictive analysis

How has the court historically ruled on this provision? What's the trend? Numbers, not guesswork.

6M+ judgments

Indian High Court judgments indexed. Inconsistent court formatting handled. Multilingual quirks in older judgments resolved.

Built for India

Indian legal text doesn't behave like Western legal text. LegalPro was built on the actual corpus — not a Western model adapted badly.

What it took

Show your work.

These are the actual hard parts of building LegalPro, not the pitch version.

01Parsing inconsistent court formatting across decades of judgments — no standard structure, mixed encodings, scanned PDFs alongside digital text.
02Building a citation graph across decades of High Court decisions — resolving ambiguous references, alternate case-number formats, and judge-name variations.
03Fine-tuning summarisation on Indian legal text specifically — Western legal models hallucinate badly on Indian HC judgments. We built training data from scratch.
04Indexing 6M+ documents into a vector store that returns in under two seconds on a query with no pre-computed answer.
Dataviz screenshot
Live
Dataviz by DemanualAI

Conversational analytics. Connect your database, ask questions in plain English from WhatsApp, Slack, or the web — and get charts, metrics, and PDF reports back in seconds.

Why it exists: every company has a single overworked data analyst with a queue three weeks deep. Dataviz removes the queue — not by replacing the analyst, but by handling the routine asks so they can focus on the hard ones.

Dataviz · features

Ask in plain English

No SQL, no dashboards, no pivot tables. Type your question the way you'd ask a colleague.

WhatsApp · Slack · Web

Works where your team already is. No new tool to adopt, no login to remember.

Charts and metrics

Auto-selects the right chart type. Renders bar, line, scatter, and tables in the response.

PDF reports

Ask for a report and get a formatted PDF — ready for the board, investor, or client.

Any database

Postgres, MySQL, BigQuery, Snowflake. Connect in minutes. No ETL required.

You own the data

Your database, your infrastructure. Dataviz reads; it doesn't store. No data leaves your environment.

§ What they prove

Not side projects.

These products aren't weekend experiments. They're how we prove what we can build. Both went from concept to production with the same small team. Both run on real infrastructure, handle real users, and break in real ways that get fixed.

If you can explain a problem clearly, we can probably build it. If we can't, we'll say so on the first call — and recommend someone we trust who can.

6M+ judgments. 99.7% uptime. 12 weeks to production.
Numbers on live infrastructure · not a demo
§ Get started

Ready to see it live?

Book a 30-minute call. One of us will walk you through the product on your data. Free, no commitment.

Book a demoSee how we work with clients