Early-career researcher applying state space models to mechanistic interpretability and anaesthesia prediction
2
papers here
2
reviewed by Evolve
2
first-author
1
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) 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 publication-count or productivity metrics are available in this snapshot, so output is left unscored. This axis would measure total publications, venue spread, and first/corresponding-author volume over time.
Impact—
Citation counts and downstream-influence measures are absent from the current data, so impact cannot be scored. This axis would measure citations, uptake of methods or findings, and influence on follow-up research.
Focus0/100
The field distribution is an exact 50/50 split between clinical machine learning (anaesthesia informatics) and mechanistic interpretability, which yields the minimum focus score of 0/100 in the radar. The two catalogued 2026 papers sit on opposite sides of this divide, though they share a Mamba/state space model technical core; concentrating on one sustained line of work would sharpen the focus profile.
Collaboration36/100
Collaboration scores 36/100, reflecting a compact working pattern: the snapshot records five distinct coauthors, with Sudiksha Chindula, Bhaskarjyoti Das, Rupa Laalasa Oruganti, Prerna Subhashchandra Kunder, and Ashwini M Joshi as recurring partners. Both reviewed papers are first-authored, one with corresponding-author status, indicating Nagaraj drives study design and writing within a small, stable team rather than a broad network.
Activity—
Longitudinal publication cadence is not captured in this snapshot, so activity is left unscored. This axis would measure recent output rate, submission consistency, and engagement across venues over consecutive years.
Reproducibility55/100
Reproducibility averages 55/100 across the two papers with recorded assessments (reproPapers = 2, reproMean = 55), both rated medium readiness. The clinical paper earns credit for solid ablations of its pharmacokinetic state space model on VitalDB, while the interpretability study is held back by uncalibrated Jaccard thresholds and a single twin-pair case study; calibrated thresholds and broader case coverage would likely lift the score.
Research summary
Rithin Nagaraj is an early-stage researcher at PES University whose reviewed output to date splits evenly between two lines of work: mechanistic interpretability of large language models and clinical machine learning for anaesthesia informatics (a 50/50 field distribution across two Evolve-catalogued arXiv papers, both from 2026, both first-authored, one also corresponding-authored). The unifying technical thread is the state space model architecture, notably Mamba. In interpretability, Nagaraj uses sparse autoencoders to compare latent concept formation between the Mamba state space model and the Pythia transformer family, reporting 99.98% feature alignment that supports the universality hypothesis — though Evolve's review flags reliance on uncalibrated Jaccard thresholds and a single twin-pair case study. On the clinical side, Nagaraj proposes pharmacokinetic state space models that ingest drug effect-site concentrations with lead-gap filtering to predict intraoperative hypotension and haemodynamic collapse on VitalDB, reporting AUROC 0.736 with solid ablations, while the mechanistic claims are judged to overreach the evidence. Collaboration is small-team: five distinct coauthors appear in the snapshot, with a radar collaboration score of 36, and reproducibility sits mid-range at 55 across the two papers, both rated medium readiness. Citation-based output, impact, and activity metrics are not yet available in this snapshot.
Collaboration pattern
Nagaraj works in a compact team: the snapshot records five distinct coauthors, and the recurring group comprises Sudiksha Chindula, Bhaskarjyoti Das, Rupa Laalasa Oruganti, Prerna Subhashchandra Kunder, and Ashwini M Joshi. On both reviewed 2026 papers Nagaraj is first author, adding corresponding-author status on the clinical paper — a lead-author pattern consistent with the modest 36/100 collaboration score, pointing to concentrated co-authorship rather than a broad network.
Research keywords
Highlights
Near-total SAE feature alignment between Mamba and Pythia
Pharmacokinetic state space model for intraoperative hypotension
One architecture, two domains: state space models across interpretability and clinical prediction
Both 2026 first-author papers apply Mamba-style state space models — one probing whether Mamba and Pythia form comparable latent concepts, the other encoding pharmacokinetic structure for perioperative prediction — producing the 0.5/0.5 dual-field footprint that defines this early-stage profile.
Timeline
- 2026Publishes first- and corresponding-author arXiv paper on pharmacokinetic state space models for unbiased prediction of haemodynamic collapse, reporting AUROC 0.736 on VitalDB.
- 2026Publishes first-author arXiv paper comparing latent concept formation in Mamba and Pythia via sparse autoencoders, reporting 99.98% feature alignment.
- 2026Both reviewed outputs land in the same year, establishing an even 0.5/0.5 split between clinical machine learning and mechanistic interpretability.
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.
- Radar scores are computed from public bibliometric data (OpenAlex-based); citation counts, h-index, and productivity totals are absent from this snapshot.
- Coverage is limited to the two Evolve-reviewed 2026 arXiv preprints; peer-review status and any earlier or concurrent publications are not recorded.
- Output, impact, and activity dimensions are marked as missing rather than scored; no estimates have been substituted.
- Profile is name-resolved from ORCID 0009-0008-9919-2466 and public bibliographic data; affiliation is limited to PES University, with no country recorded.