Susil Kumar Mohanty

Research profile

Research summary

Susil Kumar Mohanty is a researcher at the Indian Institute of Technology Jodhpur, India (ORCID 0000-0002-8643-5501), profiled from OpenAlex bibliographic data. His record is modest but clearly accelerating: an h-index of 5 and i10-index of 4, six works in the last three years, and four works recorded in 2026 against two in 2024 and a sparse tail back to 2015. OpenAlex topic history spans cryptography and data security (7 works), blockchain technology applications and security (7), and VLSI/FPGA design plus copper interconnect reliability (4 works each), while his two Evolve-reviewed 2026 papers sit squarely in AIOps — a pattern consistent with a recent pivot toward machine learning for systems. Those papers, TriFleetRCA and TriCalRAG, are arXiv preprints on which he holds a middle, non-corresponding authorship; they build single-GPU, on-premise LLM pipelines for Kubernetes root cause analysis and retrieval-augmented log-anomaly benchmarking. Evolve's verdicts credit honest, small-scale evaluation and high reproducibility readiness — a reproducibility mean of 100 across both papers — while flagging wide confidence intervals and same-pool retrieval risk. His collaboration footprint is compact: two distinct coauthors, Rohit Patel and Jeenal Chaudhary. Citation-based output and impact scores are unavailable in this snapshot, and the focus score of 0 reflects topic dispersion rather than absence of work.

Research keywords

Highlights

• Single-GPU on-premise LLM root cause analysis for Kubernetes

  TriFleetRCA (2026, arXiv) delivers on-premise root cause analysis for Kubernetes on a single GPU, reaching 0.85–0.95 over 4 injected fault types (n=100 runs). Evolve's verdict calls the evaluation honest but small-scale, with wide confidence intervals and one toy attack — and explicitly reproducible.

• Three-strategy RAG benchmark for LLM-based AIOps diagnosis

  TriCalRAG (2026, arXiv) benchmarks three retrieval-augmented strategies and shows RAG stabilizes zero-shot-degenerate LLM log-anomaly detection on one GPU, with good calibration analysis; Evolve flags small scale and a same-pool retrieval risk of score inflation.

• A reproducibility-first preprint record

  Both reviewed papers rate high on reproducibility readiness, yielding a reproducibility mean of 100 over 2 assessed papers — the strongest axis in this profile, tied to consistently single-GPU, on-premise experimental setups.

Timeline

1. 2015Earliest publication year recorded in the OpenAlex snapshot, with 1 work.
2. 2021Most productive early year on record, with 3 works, amid a generally sparse 2015–2023 output tail.
3. 20242 works recorded, marking the ramp-up toward the recent AIOps-focused output.
4. 2026Strongest year on record with 4 works, including both Evolve-reviewed AIOps preprints, TriFleetRCA and TriCalRAG.

Papers in the Evolve catalog

• 2026-09-20

  TriFleetRCA: On-Premise LLM Root Cause Analysis for Kubernetes

• 2026-09-13

  TriCalRAG: A Three-Strategy, Retrieval-Augmented Benchmark for On-Premise LLM-Based Root Cause Analysis in AIOps

Frequent co-authors