Oluwatobi Olabiyi

Nvidia (United States) · United States

Unclaimed · AI-structured profileOpenAlex

Nvidia researcher advancing full-duplex speech-to-speech models and tool-calling voice agents

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Not scored: the record reports h-index and i10-index of 0 with no reliable output counts, so productivity cannot be assessed. This axis would measure publication volume and its trend once the record matures.

Impact—

Not scored: citation indicators (h-index, i10-index, per-paper citations) are effectively unavailable for a record spanning only 2025-2026. This axis would measure scholarly uptake and real-world adoption of the models once citations accumulate.

Focus0/100

Scores 0/100, which should be read against the raw distribution: the field split is exactly even (0.5/0.5) between 'speech processing and conversational AI' and 'speech processing / spoken dialogue systems (cs.CL)' — two adjacent labels for the same domain — and topic counts (Topic Modeling 6, Advanced Neural Network Applications 5, NLP Techniques 4, two speech topics at 4 each) confirm a speech-centric record. The 0 likely reflects label fragmentation rather than genuine breadth.

Collaboration83/100

Scores 83/100. The OpenAlex record lists 64 distinct co-authors across the recent three-year window alongside 10 works, a ratio that points to large-team industrial writing rather than solo research. Recurring partners include Edresson Casanova, Zhehuai Chen, Jagadeesh Balam, Elena Rastorgueva, Nikhil Srihari, Slyne Deng, Travis Bartley, Viet Anh Trinh, Sanjay Chauhan and Chen Chen; both reviewed papers are multi-author efforts.

Activity70/100

Scores 70/100. countsByYear shows 1 work in 2025 and 9 in 2026, and the recent three-year window holds 10 works — so nearly all recorded output lands in a single year, a steep ramp typical of a newly visible or newly intensive phase. Both Evolve-reviewed papers are 2026 arXiv preprints, consistent with that acceleration.

Reproducibility38/100

Scores 38/100, the mean over 2 Evolve-reviewed papers (reproMean 38, reproPapers 2). NemotronLabs VoiceChat is rated medium readiness, while the frontend-backend tool-calling paper is rated low: Evolve flags untested design claims, missing significance testing, and weak argument accuracy (42.2%, Pass@1 33.0%). Artifacts may exist, but independent verification coverage is incomplete.

Research summary

Oluwatobi Olabiyi is a researcher at Nvidia (United States) working in speech processing and conversational AI; this profile is name-resolved from OpenAlex bibliographic data, with no ORCID supplied. The record shows 10 works in the recent three-year window, sharply concentrated in 2026 (9 works versus 1 in 2025), and a large collaboration footprint of 64 distinct co-authors, yielding a collaboration score of 83 and an activity score of 70. Evolve has reviewed two 2026 arXiv papers, both with the researcher in a middle author position. The first, NemotronLabs VoiceChat, is credited as the first open full-duplex speech-to-speech model with native tool calling, with strong turn-taking but weak argument accuracy (42.2%; Pass@1 33.0%); the second proposes a frontend-backend architecture that adds tool calls to a full-duplex speech frontend via token-based delegation, with strong benchmark results but no significance testing. Topic counts place the record in Topic Modeling (6), Advanced Neural Network Applications (5), Natural Language Processing Techniques (4), Speech Recognition and Synthesis (4), and Speech and Audio Processing (4); the field distribution splits evenly (0.5/0.5) between speech/conversational AI and spoken dialogue systems. The radar reproducibility score is 38, the mean over those 2 reviewed papers. Output and impact remain unscored: h-index and i10-index read 0 and citation data is insufficient for a record that only reaches back to 2025.

Collaboration pattern

The OpenAlex record lists 64 distinct co-authors over the recent three-year window, and the radar collaboration score is 83. Recurring names — Edresson Casanova, Zhehuai Chen, Jagadeesh Balam, Elena Rastorgueva, Nikhil Srihari, Slyne Deng, Travis Bartley, Viet Anh Trinh, Sanjay Chauhan and Chen Chen — suggest a stable large-team pattern typical of an industrial speech lab, with both reviewed papers written by multi-author groups.

Research keywords

full-duplex speech-to-speechtool callingspoken dialogue systemsturn-takingbarge-instreaming ASRspeech recognition and synthesisspeech and audio processingvoice agentsnatural language processingtopic modelingneural network applications

Highlights

First open full-duplex speech-to-speech model with native tool calling
Evolve's review of NemotronLabs VoiceChat credits it as the first open full-duplex S2S model with native tool calling, with strong turn-taking, barge-in handling and streaming ASR — while flagging weak argument accuracy (42.2%, Pass@1 33.0%) and untested design claims; reproducibility readiness is medium.
Token-based delegation brings tool calls into full-duplex speech frontends
The frontend-backend architecture paper adds tool calling to a duplex speech frontend via a delegation token; Evolve notes strong benchmark results, but no significance testing and low reproducibility readiness.
Reproducibility is the open gap across both reviewed papers
Across the two Evolve-reviewed papers, the mean reproducibility readiness is 38/100 (one medium, one low), with untested design claims and missing significance testing cited — the clearest axis for improvement.

Timeline

2026Publication tempo peaks: OpenAlex countsByYear logs 9 works in 2026 versus 1 in 2025, so nearly the entire recent three-year output of 10 works lands in this single year.
2026Co-authors NemotronLabs VoiceChat, positioned as the first open full-duplex speech-to-speech model with native tool calling, in a middle author position.
2026Co-authors the frontend-backend architecture paper, which uses token-based delegation to add tool calls to full-duplex speech models, also in a middle author position.

Papers in the Evolve catalog

Frequent co-authors

Edresson Casanova×2 Zhehuai Chen×2 Slyne Deng×2 Elena Rastorgueva×2 Nikhil Srihari×2 Viet Anh Trinh×2 Jagadeesh Balam×1 Travis Bartley×1 Edresson Casanova×1 Sanjay Chauhan×1 Chen Chen×1 Zhehuai Chen×1

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

Bibliometric data comes from OpenAlex only; the profile is name-resolved from public bibliographic data (no ORCID provided), so homonymous records cannot be fully excluded.
h-index and i10-index are reported as 0 and per-paper citation counts are unavailable, so the output and impact dimensions are left unscored rather than estimated.
Evolve has reviewed only 2 papers, both 2026 arXiv preprints and not yet peer-reviewed; both list this researcher as a middle author rather than first or corresponding.
The reproducibility score (38) is a mean over those 2 reviewed papers only and may not reflect the broader research portfolio.