Early-career researcher spanning LLM agent security, governance-as-code compliance, and cost-optimal LLM routing
3
papers here
3
reviewed by Evolve
3
first-author
0
corresponding
—
works · OpenAlex
4
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-09) 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—
This axis would measure publication volume and productivity relative to career stage and field norms, but output metrics are not yet computed for this profile. The reviewed record comprises only three 2026 arXiv preprints, too small a base to score fairly.
Impact—
This axis would measure citation-based influence such as citations per paper and field-normalized impact. It is unscored: visible signals are limited (h-index 4, i10-index 0), reflecting an early-stage citation footprint rather than a mature impact record.
Focus0/100
Focus scores 0 because output is spread almost evenly across three areas: per the field distribution, LLM agent security, AI governance/compliance engineering, and LLM inference optimization each account for about one-third. This is a deliberately multi-front, applications-oriented agenda rather than a concentrated one, and OpenAlex topics (AI and big data impact on business and society: 9; digital transformation: 6) show an even broader applied span.
Collaboration14/100
Collaboration scores 14, near the bottom of the scale: the raw record lists just 1 distinct coauthor, recurring collaborator Sourav Nandy. All three reviewed papers are first-authored with no corresponding role, suggesting an independent, small-team working style. A near-solo model gives full control of the research agenda but limits the co-authorship network, external validation, and scale that broader teams provide.
Activity84/100
Activity scores 84, the strongest axis. OpenAlex counts by year show steady early output (2 works in 2021, 2 in 2022, 3 in 2023, 2 in 2024) followed by a sharp acceleration to 8 works in 2025 and 7 already in 2026, with 17 works over the last three years — a clearly intensifying cadence centered on 2026 arXiv preprints.
Reproducibility20/100
Reproducibility scores 20, the mean across 3 assessed papers, each rated low readiness. Evolve's verdicts flag concrete gaps: the AgentRouter evaluation lacks significance tests, the multi-agent injection defense is validated on a single system without confidence intervals, and the Governance-as-Code pipelines were tested only on two private case studies with no code release.
Research summary
Rudrendu Kumar Paul is an early-career researcher whose profile is name-resolved from public OpenAlex bibliographic data (ORCID 0009-0008-0141-4690); no institution or country is recorded. The Evolve catalog holds three first-authored 2026 arXiv preprints spanning three equally weighted directions — LLM agent security, AI governance/compliance engineering, and LLM inference optimization for agentic systems — matching the raw field distribution of one-third each. The signature is fast, systems-level problem framing: a routing scheme reporting 71.8% cost reduction at 97.3% quality (AgentRouter); a threat model of 14 multi-agent prompt-injection vectors with 4 defenses cutting attack success from 31.2% to 4.2%; and 43 executable Rego checks translating EU AI Act Articles 8-15 into CI/CD compliance gates. Activity is the strongest radar axis (84): yearly works rose from 2 (2021, 2022, 2024) and 3 (2023) to 8 in 2025 and 7 already in 2026, with 17 works over the last three years. Citation mass remains thin (h-index 4, i10-index 0), so output and impact are left unscored. Collaboration is minimal (score 14; 1 distinct coauthor, Sourav Nandy) and reproducibility is weak (20): all three reviewed papers are rated low readiness, with missing significance tests, single-system evaluation, and no code release.
Collaboration pattern
The record shows an unusually independent pattern: 1 distinct coauthor in the raw data — recurring collaborator Sourav Nandy — and all three Evolve-reviewed papers are first-authored with no corresponding role. There is no evidence of large-lab or cross-institutional teams in this snapshot. Future collaboration potential lies in connecting the security, governance, and systems strands with groups that can supply shared benchmarks and released artifacts.
Research keywords
Highlights
Step-level routing cuts agent LLM costs
Threat model for multi-agent prompt injection
EU AI Act requirements as executable code
Timeline
- 2021Earliest works in the OpenAlex record: 2 works published this year.
- 20233 works recorded, the highest annual count before the recent surge.
- 2025Output accelerates sharply to 8 works, up from 2 in 2024.
- 2026Most intensive year so far: 7 works recorded, including three first-authored arXiv preprints reviewed by Evolve — AgentRouter, the multi-agent prompt-injection threat model, and Governance-as-Code.
Papers in the Evolve catalog
- first author
AgentRouter: Heterogeneous Model Routing for Cost-Optimal Multi-Step Agentic Workflows
- first author
Beyond Single-Model Injection: A Threat Model and Defense Architecture for Prompt Injection in Multi-Agent Systems
- first author
Governance-as-Code: Translating EU AI Act Technical Requirements into Executable Compliance Pipelines for Generative AI Systems
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 signals come from OpenAlex only; institution and country are absent from this snapshot, and the profile is name-resolved from public bibliographic data (ORCID 0009-0008-0141-4690).
- Output and impact radar axes are unscored because underlying data is insufficient; beyond the reported h-index (4) and i10-index (0), per-paper citation counts are not available.
- Only three 2026 arXiv preprints carry Evolve review verdicts and reproducibility ratings; the other works counted in OpenAlex yearly totals are not individually assessed.
- OpenAlex topic labels (e.g., supply chain and logistics) point to broader earlier work outside the three reviewed LLM papers; those earlier works are not reviewed here.