Sai Karthik Kosuri
University of Pennsylvania · United States
Penn NLP researcher: human-in-the-loop evidence synthesis and speech/language markers for mental health
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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—
No output volume is scored yet: the snapshot lists only the two 2026 arXiv records, with h-index and i10-index at 0. This axis would measure total works, annual output rate, and venue spread once a fuller corpus is linked.
Impact—
Impact is not scored: h-index and i10-index are both 0 and no citation counts appear in the snapshot. This axis would track citations to the two 2026 papers, downstream adoption, and influence once the work has time to accrue citations.
Focus0/100
Focus scores 0 because the field distribution is an even 0.5/0.5 split between computational psycholinguistics/digital mental health and clinical natural language processing/evidence synthesis, with no dominant field yet. The five indexed topics each carry a count of 1, reinforcing the balanced picture. Read as dual-track rather than scattered: both cataloged papers apply language-centered methods — extraction pipelines and speech/LIWC analysis — to clinical and mental-health questions.
Collaboration53/100
The collaboration score of 53 rests on 13 distinct coauthors in the OpenAlex snapshot, with 12 recurring partners. The roster plausibly spans both research strands: computing and NLP colleagues alongside clinical and evidence-synthesis collaborators such as Alonso Carrasco-Labra, Michael Glick, Thomas K. M. Cudjoe, and Matthew Lee Smith. Kosuri is first author on one cataloged paper and a middle author on the other, suggesting both leading and supporting roles.
Activity32/100
Activity scores 32, consistent with a very early career stage. The countsByYear record contains a single entry — 2026 with 2 works — and recent3yWorks is likewise 2, meaning all indexed output falls within the current year and both items are arXiv preprints. No earlier publication years appear in the snapshot, so the score reflects a short observable history rather than a slowdown; future cycles will show whether this rate is sustained.
Reproducibility38/100
Reproducibility averages 38 across the 2 papers Evolve has assessed (reproMean 38, reproPapers 2). The evidence-synthesis system paper is rated medium readiness, while the loneliness study is rated low. Evolve's verdicts flag concrete limitations: calibration and judge circularity constrain the extraction claims, and interview-level correlations may inflate significance in the loneliness analysis. The candid self-reported evaluation is a positive sign, but code and data availability remain unconfirmed in the snapshot.
Research summary
Sai Karthik Kosuri is an early-career researcher at the University of Pennsylvania working at the intersection of clinical natural language processing and computational psycholinguistics for mental health. The OpenAlex-linked profile is name-resolved from public bibliographic data and currently rests on two 2026 arXiv papers, both reviewed in Evolve's catalog. The first line of work, evidence synthesis, is represented by EviStreams, a first-authored system for human-in-the-loop AI data extraction in medical systematic reviews; Evolve's verdict credits the paper's claim that field specifications — rather than model choice — drive extraction quality, while noting calibration and judge-circularity limits. The second line, digital mental health, contributes to a multimodal analysis of speech and language as predictors of loneliness in 310 older adults, reporting a multimodal correlation of r=0.298, with Evolve cautioning that interview-level correlations may inflate significance. Bibliometric signals are correspondingly early-stage: h-index and i10-index are both 0, countsByYear contains only 2026 with 2 works, and the field distribution splits evenly (0.5/0.5) between the two strands, yielding a focus score of 0. Collaboration is the strongest scored dimension at 53, with 13 distinct coauthors spanning NLP and clinical medicine. Reproducibility averages 38 across the two assessed papers. The profile should be read as a snapshot of an emerging researcher rather than an established record.
Collaboration pattern
The OpenAlex snapshot records 13 distinct coauthors, with 12 recurring partners including Chris Callison-Burch, Sharath Chandra Guntuku, Alonso Carrasco-Labra, Michael Glick, Thomas K. M. Cudjoe, Matthew Lee Smith, Neil K. R. Sehgal, Vinmay Khandode, Adam Greene, Elif Alpoge, Elana Duffy, and Ankita Shashikant Bhosale. The roster plausibly spans both cataloged fields — clinical NLP/evidence synthesis and computational psycholinguistics/digital mental health — and Kosuri appears as first author on one paper and a middle author on the other.
Research keywords
Highlights
First-authored system for human-in-the-loop extraction in medical systematic reviews
Multimodal speech and language markers of loneliness in 310 older adults
Balanced dual-strand research profile
The OpenAlex field distribution splits evenly (0.5/0.5) between computational psycholinguistics/digital mental health and clinical NLP/evidence synthesis; five indexed topics each carry a count of 1, and both cataloged papers were posted in 2026.
Timeline
- 2026First-authored EviStreams posted to arXiv: human-in-the-loop AI data extraction for systematic reviews in medicine; assessed by Evolve with medium reproducibility readiness.
- 2026Co-authored multimodal study of speech and language predictors of loneliness in 310 older adults (multimodal r=0.298); Evolve rates reproducibility readiness low.
- 2026OpenAlex snapshot records 2 works in countsByYear for 2026 (recent3yWorks = 2), with 13 distinct coauthors and an even 0.5/0.5 field split across the two strands.
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 metrics are derived from OpenAlex only (identity: openalex); no ORCID was provided, so the profile is name-resolved from public bibliographic data.
- The cataloged corpus is small and recent: two 2026 arXiv preprints; h-index and i10-index are both 0, so output and impact dimensions are unscored (dimMissing).
- Reproducibility reflects Evolve assessments of only 2 papers (mean 38); readiness ratings are medium and low respectively.
- Citation counts, venue acceptance, funding, and career history are absent from the input data and are not estimated.