Speech AI researcher at Nvidia working on full-duplex speech-to-speech models and voice-agent tool calling
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
11
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 yet scored: this snapshot lacks complete per-work publication records, so the output axis — which would measure research volume and consistency — is left unscored; a 10 h-index and 11 i10-index are reported without a derived score.
Impact—
Not yet scored: per-paper citation counts and venue-level uptake are not available in this snapshot, so the impact axis — which would measure citation influence and downstream adoption of the work — is left unscored rather than estimated.
Focus8/100
Scores 100: the field distribution assigns the entire classified record to a single field, speech processing and conversational AI, and topic counts confirm the concentration — Speech Recognition and Synthesis (55), Natural Language Processing Techniques (28), Speech and Audio Processing (21), Music and Audio Processing (15), and Speech and dialogue systems (13) all sit within speech and language technology.
Collaboration86/100
Scores 62 on a base of 21 distinct coauthors, with a recurring twelve-name core — including Ke Hu, Elena Rastorgueva, Piotr Zelasko, Yifan Peng, Zhehuai Chen, and Boris Ginsburg — typical of a large industrial speech team. Both cataloged 2026 papers are middle-author, non-corresponding contributions, consistent with team-based production rather than solo-led projects.
Activity97/100
Scores 100 on sustained output: the year-by-year record runs from 2 works in 2019 to a peak of 14 in 2022, dips to 8 (2023) and 6 (2024), then rebounds to 13 in 2025 with 11 already in 2026, and 30 works fall within the trailing three years.
Reproducibility43/100
Scores 55, drawn from one scored paper in the catalog. Evolve's read on the two 2026 papers flags a consistent pattern: the tool-call architecture paper reports strong benchmark results with no significance testing (low reproducibility readiness), while the streaming-transcription paper is judged sensible but statistically thin (medium).
Research summary
Edresson Casanova is a speech AI researcher at Nvidia (United Kingdom) whose work centers on full-duplex speech-to-speech models and spoken dialogue systems. Bibliometric coverage resolved from OpenAlex shows a tightly focused profile: the field distribution attributes all classified output to speech processing and conversational AI, and topic counts are led by Speech Recognition and Synthesis (55), Natural Language Processing Techniques (28), and Speech and Audio Processing (21). Activity is the standout axis (100/100): the year-by-year record rises from 2 works in 2019 to a 2022 peak of 14, dips through 2023-2024 (8 and 6), and rebounds with 13 works in 2025 and 11 already in 2026, with 30 works in the last three years. Collaboration scores 62, built on 21 distinct coauthors and a recurring core including Ke Hu, Piotr Zelasko, and Boris Ginsburg. Two 2026 arXiv papers in the Evolve catalog — tool-call delegation for full-duplex speech models and streaming user transcription via a parallel ASR head (10.21% WER) — show a researcher shipping interactive speech-system components, though Evolve flags thin statistics and low-to-medium reproducibility readiness. An h-index of 10 and i10-index of 11 are reported, but citation-derived output and impact axes remain unscored.
Collaboration pattern
Collaboration is team-based: 21 distinct coauthors appear in the record, and twelve recur often enough to be flagged — Ke Hu, Elena Rastorgueva, Nourchene Ferchichi, Ankita Pasad, Chen Chen, Nithin Rao Koluguri, Piotr Zelasko, Yifan Peng, Hainan Xu, Zhehuai Chen, Boris Ginsburg, and Slyne Deng. Both cataloged 2026 papers are middle-authored and non-corresponding, consistent with contributing inside a large Nvidia speech group rather than leading standalone efforts.
Research keywords
Highlights
Tool-call delegation for full-duplex speech models
Streaming user transcription inside a duplex S2S model
Timeline
- 2019Earliest works in the bibliometric record appear, with 2 publications
- 2022Annual output peaks at 14 works after climbing through 12 in 2021
- 2023Output dips across 2023-2024, with 8 and 6 works respectively
- 2025Activity rebounds to 13 works in the year
- 202611 works already recorded; two full-duplex speech papers (tool-call delegation and streaming user transcription) enter the Evolve catalog
Papers in the Evolve catalog
Frequent co-authors
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 figures are resolved from an OpenAlex identity record; the profile is name-resolved from public bibliographic data, not verified biography.
- Output and impact radar axes are unscored because per-paper citation coverage in this snapshot is insufficient.
- The reproducibility score rests on a single scored paper (reproPapers: 1), so the mean of 55 is thin evidence.
- Only two papers have been reviewed in the Evolve catalog; both are arXiv preprints with middle-author, non-corresponding positions.