Product 1 of 5 · Research copilot

1MentraResearch copilot

Answers with sources, actions with approval

Mentra reads the literature, patents and your own notebooks, explains why a candidate ranked where it did, and drafts the next campaign. It never launches work without a named person approving it.

Abstract

Role
Research copilot
Loop step
1 · frame the question
Grounding
Literature, patents, your data
Built on
NeMo, NIM, TensorRT-LLM
Example · Mentra sessionExample
Which halide substitutions in Li₆PS₅Cl have been reported to raise room-temperature conductivity, and which did our last screen keep?
Two families appear repeatedly in the literature: Br-for-Cl mixing and Cl-rich compositions12. Your round 1 screen kept 9 Br-mixed candidates within 25 meV/atom; 3 of them also clear 1 mS/cm in MD3. None of the I-substituted candidates passed the hull filter.
Proposed action: start round 2 on Cl-rich Br mixes, 180 GPU-hAwaiting approval
  1. [1] Literature source (placeholder citation)
  2. [2] Patent family (placeholder citation)
  3. [3] Campaign sse-li-argyrodite-r1, results table, rows 14–22
Example. An illustrative exchange. Citations are placeholders; in the product each one links to the exact passage or result row.

01 · Capabilities

What Mentra does

Grounded answers

Retrieval with citations

Every claim links to a passage in a paper, a patent or a row in your results. If there is no source, Mentra says so.

Explain a ranking

Why this candidate

Breaks a ranking into hull distance, predicted property, uncertainty and the filters that applied.

Draft a campaign

From question to config

Turns a question into a campaign file with search space, filters and a GPU budget for you to edit.

Approval-gated actions

Nothing runs alone

Proposes screens or lab batches; a named person approves each one in the console.

Your archive

Old notebooks count

Indexes past DFT runs, reports and lab notes in your tenant, so earlier work is not repeated.

Guardrails

IP boundaries

Keeps answers inside the documents a user is allowed to see, and logs every retrieval.

02 · Method

How Mentra works

  1. 1

    Ask

    A question in plain language, scoped to a program.

    Input TextTime Seconds
  2. 2

    Retrieve

    Search literature, patents and tenant data; rank passages by relevance and recency.

    Uses NeMo Retriever
  3. 3

    Answer with citations

    Compose an answer only from retrieved passages; mark gaps explicitly.

    Uses NIM, Guardrails
  4. 4

    Propose an action

    Draft a campaign or batch with its GPU cost for approval.

    Gate Named approver

03 · Specification

Specification

First-release scope. Items marked planned are not in the first release.
ItemDetail
InputsQuestions; literature and patent corpora (licensing per source); tenant documents and results
OutputsCited answers, ranking explanations, draft campaign files, approval requests
ActionsStart a Dynara screen, propose an Inventa batch, route to Synthex: all approval-gated
ModelsNeMo-based LLM with retrieval and guardrails, served with NIM
Voice and multimodalSpeech in the lab and reasoning over spectra Planned
AuditEvery query, retrieval and action logged with user and time

04 · Place in the loop

Where Mentra sits

Mentra is the front door. It frames the question at the start of the loop and explains each result as it comes back, so a program lead can follow a campaign without reading every table.

1MentraResearch copilot

Reads the literature and your notebooks, frames the question and drafts the campaign.

2InventaDesign + active learning

Turns the target into candidates and picks the next batch by expected information.

3EquivaInteratomic potentials

Predicts energies, forces and stresses at near-DFT accuracy, with an uncertainty for each.

4DynaraGPU simulation

Relaxes, scores against the hull and runs GPU molecular dynamics on the survivors.

5SynthexSelf-driving lab

Sends the short list to partner or in-house labs and brings measurements back.

Return pathSynthex measurements → fine-tune Equiva → better predictions in the next round.

05 · Questions

Questions about Mentra

Does Mentra make things up?

It is built to answer only from retrieved sources and to say when it has none. It can still misread a source, which is why every claim carries a link you can check.

Which literature can it read?

Open-access literature to start; licensed publisher and patent corpora depend on agreements we have not signed yet. Your own documents are always available to your tenant.

Can it run without approval for small jobs?

Not in the first release. We may add per-team budgets for small screens later, if pilot partners ask for it.

Next step

Test Mentra on one of your chemistries

The 2027 pilot pairs one of your programs with our scientists for twelve weeks. You bring a target and reference data; we bring the GPUs and the models.