Asteria Deep Research: balancing speed and rigor in R&D state-of-the-art

A promising concept rejected in technical committee. A direction pursued for months, before discovering a competitor had already published on it. If you're an R&D engineer and these situations sound familiar, the cause is usually not a lack of ideas. It's a state of the art done too fast, or not done at all.

At Asteria, exploration, design and analysis each have a dedicated tool. Deep Research, available since recently, is built for the last one: understanding a market, a technology or a mechanism before you commit.

What is Deep Research, and how does it de-risk a state-of-the-art R&D?

Deep Research is Asteria's new in-depth research mode. It produces a structured, state-of-the-art on an R&D question in 15 to 30 minutes.

In practice: you launch a search, work on something else, and come back to a complete report that clears the ground before you start designing.

Each section is then researched in two steps. Asteria first queries its proprietary database to identify the key mechanisms and information. This first search is systematically complemented by a web search, which refines and adapts the results, for example to your company's sector.

The two searches are then assembled into one complete, coherent report.

Unlike Chat or Project, which require ongoing dialogue with you, Deep Research runs autonomously. You launch it, it runs, you get the result.

Why isn't keyword search enough?

An engineering problem isn't solved by searching for terms. It's solved by reasoning through function: what does the system need, and which strategies fulfill that function?

Say you're working on a surface that needs to stay clean without chemical treatment.

Searched by keyword, the query returns what's already been written extensively on the topic, and nothing more.

Searched by function, how to prevent contaminant adhesion on a surface, it opens up a much wider field: from the microstructures of certain leaves to the properties of animal skin, along with the physico-chemical principles behind them.

A general-purpose search tool sweeps the web indiscriminately. It doesn't distinguish a reference source from surface-level content, and it has no framework for breaking down an engineering problem. It gives the impression the topic is covered. It isn't.

What makes this database different: the function-based knowledge graph.

Deep Research runs on a proprietary bio-inspired engineering methodology, not a generic language model querying the web at random.

Its knowledge base holds over 680,000 biological mechanisms, 1.3 million scientific publications and 300,000 bio-inspired patents, indexed by function rather than by keyword.

This functional indexing links each mechanism to the papers that explain it and the patents that have already applied it.

That's what surfaces functional analogies and a history of technology transposition that an undifferentiated web search can't reconstruct.

It's also why the search takes time. Depth requires more than a few seconds of computation.

When should you use Deep Research?

Four use cases come up most often with our users:

- Technology watch: mapping existing bio-inspired technologies on a given topic, across all maturity levels.

- Market study: assessing strategic and economic opportunities around a theme.

- Cross-functional analysis: combining several of these angles in a single report.

- Scientific and academic literature review: identifying the actors, institutions and consensus around a bio-inspired mechanism.

A solid state-of-the-art makes for a defensible decision.

De-risking R&D innovation is usually associated with the last steps: testing, prototyping, validation. The first lever, though, comes much earlier. A project starts on solid ground when you know, from day one, what you're building on and what you still don't know.

Rebuilding the state of the art isn't saving time on a tedious step. It's addressing risk while you can still correct it without starting over.

Interested in seeing Deep Research in more detail?‍

Book a demo with one of our experts.