Skip to content
BiVelio joins the NVIDIA Inception Program NVIDIA Inception Program BiVelio is part of the ElevenLabs Grants Program ElevenLabs Grants

Research by BiVelio

We teach the algorithms we build and the models we use to measure risk.

Retrieval, agents, knowledge graphs and orchestration — and now quantitative finance, with the leveraged ETF Modeler. No hype: the why, the how and the math.

Open the Modeler Read the articles
articles in the library
18
algorithms in production
2
daily data since
1928
futures per simulation
2,000

Our mission

Knowledge that isn't shared doesn't compound.

We built BiVelio by running real companies. Every algorithm we design comes from a concrete problem, not a paper. We believe explaining it in the open is the most honest way to make progress.

In the open

We publish the reasoning, not just the result. Transparency is a requirement, not a marketing gesture.

With rigor

Math, code and real measurements. If we claim something, we show how we know.

That compounds

Each article builds on the last. Knowledge that accumulates is what creates long-term advantage.

To share

We write to teach. If an idea helps another team build better, it has done its job.

Line 1 · In production

Optimization before the model, not inside it

We don't replace OpenAI, Anthropic, DeepSeek or Gemini. We sit in front of them: we better prepare what you send them, so you spend less and expose less.

All algorithms

BV-PRGA · In production

BiVelio Privacy Gateway

An intermediate layer that holds back sensitive information before sending your prompt, document or file to any AI API.

Learn more

Line 2 · Quant finance

What leveraged ETFs really earn, why, and how much of it is luck

Using daily data since 1928, we study how 2x and 3x funds behave, which rules improve them and how much of what they earn can't be told apart from chance. Every article accounts for the search behind it, and confirmatory studies preregister their hypotheses before testing them.

first year of daily data
1928
articles in the series
6
paths per simulation
2,000
Open the Modeler Read the series Play with the numbers

A workstation for simulating markets

Pick a world — reshuffled history, volatility that clusters, more frequent crashes or the real crises — and compare the 3x, the 200-day moving average or half Kelly across thousands of possible futures.

Median wealth per euro invested · 1,000 paths

Latest articles in the series, on Substack (in Spanish)

Loading articles…

Line 3 · The library

Research articles

Browse by topic. Each piece is self-contained: context, intuition, formalization and results.

The whole library

Featured · Agents

Why your RAG fails and the graph recovers double

An agent without context is just a model. Classic vector retrieval treats knowledge as a bag of independent chunks — which is why it fails on the questions that matter most to a company. We walk through the real problems RAG developers hit (with data), why graph structure wins where vectors fall short, how the landscape stacks up (Pinecone, LangChain, LlamaIndex, Neo4j, Microsoft GraphRAG), and how we plan to be the best at the context that actually counts: a company's operation.

Read the article
Algorithms BV-PRGA: inside the privacy gateway September 2, 2026 Algorithms BV-SALA: inside the LLM cost layer September 2, 2026 Systems 40% of agentic AI dies without an Autonomy Rate July 2, 2026 Agents 95% of AI pilots fail: it's not the model July 2, 2026 Systems A single agent fails: governed AI in 3 layers July 2, 2026 Foundations Adding a copilot is not an operating model July 2, 2026 Systems AI agents: it's not the model, it's the architecture July 2, 2026 Foundations AI Workforce OS: not a suite, but an operating model July 2, 2026 Systems Buying more software won't make you an autonomous company July 2, 2026 Agents How to govern AI agents: authority, not output July 2, 2026 Foundations Human-in-the-loop: 5 components, not an approval button July 2, 2026 Foundations RPA automates tasks. A GPOS governs operations July 2, 2026 Agents RPA vs workflow automation vs AI agents: layers, not rivals July 2, 2026 Foundations Your ungoverned AI agent is a liability, not a milestone July 2, 2026 Retrieval Your RAG's problem isn't the model, it's the order June 20, 2026

In the classroom and on stage

Teaching and talks

Universities and platforms where we teach or have presented our work.

Universitat d'Andorra Teaching · Universitat d'Andorra uda.ad

Papers and talks

Platforms

Substack

Get every article in your inbox.

The series is published on Substack, in Spanish. Subscribe to get each article as soon as it comes out; you can unsubscribe at any time.

Subscribe on Substack