About Vitreon Legal

1.Our Mission

Vitreon Legal exists to democratize access to legal research through artificial intelligence. Every lawyer deserves the same quality of legal research regardless of firm size, budget, or geography. A solo practitioner in Brno should have the same research capabilities as a Magic Circle firm in London.

We build AI systems that deliver precise, source-grounded answers from statutes and court decisions. Every answer cites the exact page and clause. No hallucinations. No unsourced claims. Just verifiable legal research at the speed of thought.

2.Founder

Vitreon Legal was founded by Viacheslav “Neon” Ivannikov, a solo founder based in Prague, Czech Republic. The platform was born from firsthand frustration with expensive, clunky legal databases that locked critical legal knowledge behind enterprise paywalls and outdated interfaces.

What started as an entry in the ARLC 2026 competition evolved into a full platform serving legal professionals across four jurisdictions. The core conviction remains the same: AI should make legal research more accessible, not more opaque.

3.The Story

In early 2026, the Agentic RAG Legal Challenge (ARLC 2026) was announced as part of Dubai AI Week — an international competition with a $32,000 prize pool, challenging teams to build AI systems capable of answering complex legal questions from 300+ DIFC legal documents.

Vitreon entered the competition under the team name “Neon Team”. Competing against 80 teams from around the world, Neon Team placed 4th overall, scoring 0.958 on the warmup round (1st place) and 0.719 on the finals.

The competition proved that the underlying retrieval and reasoning architecture could compete at the highest level. After ARLC, the system was expanded to cover Czech law — 295,000+ court decisions and 6,800+ statutes — along with UK and Australian jurisdictions. Vitreon Legal launched as a commercial platform in 2026.

4.ARLC 2026 Competition Results

The Agentic RAG Legal Challenge 2026 was organized during Dubai AI Week with a $32,000 prize pool. 80 teams competed to build the most accurate legal question-answering system over 300+ DIFC legal documents.

Team “Neon Team” on the ARLC leaderboard is Vitreon Legal. The same retrieval pipeline that achieved these scores now powers the production platform.

5.Technology

Vitreon Legal's retrieval pipeline is built on custom-trained embedding models and a hybrid search architecture that consistently outperforms published state-of-the-art systems on independent benchmarks:

• GaRAGe benchmark (ACL 2025): 0.824 RAF score, +36% above the published SOTA of 0.607.
• LEXam Open EN (ICLR 2026): 0.691, +21% above the Claude 3.7-S baseline of 0.572.
• Legal RAG Bench: 0.860 retrieval accuracy.
• 100% citation coverage: every answer cites the exact page and clause from the source document.

The system uses hybrid BM25 + vector search with Reciprocal Rank Fusion, asymmetric embedding via Qwen3-Embedding-8B, and cross-encoder reranking. The Czech legal corpus includes 295,000+ court decisions and 6,800+ statutes.

For detailed benchmark methodology and results, see the Benchmarks page.

6.Contact

We are always happy to hear from legal professionals, researchers, and potential partners.