Parishruthi Ganesh is a Ph.D. student in Computer Science and Software Engineering at Auburn University working on large language models, NLP and computer vision. Her first-author work includes a systematic zero-shot evaluation of 41 open-weight language models and a controlled study of interaction representations for early violence detection in video.https://parishruthiganesh.github.io/en-ushttps://parishruthiganesh.github.io/hackathons#speed-august-ai-thinktrace/https://parishruthiganesh.github.io/hackathons#speed-august-ai-thinktrace/Don't just mark the wrong answer. ThinkTrace AI reads each student's written reasoning, groups the class by how it is confused, and keeps teaching until the misconception is demonstrably gone.Sat, 29 Aug 2026 00:00:00 GMTHackathonhttps://parishruthiganesh.github.io/hackathons#all-things-agentic-labguard/https://parishruthiganesh.github.io/hackathons#all-things-agentic-labguard/An autonomous research agent that challenges a scientific claim, runs the experiments that could disprove it, repairs the runs that break, and returns an evidence-backed verdict with every check shown.Sat, 29 Aug 2026 00:00:00 GMTHackathonhttps://parishruthiganesh.github.io/hackathons#proof-of-possible-2026-splitshield/https://parishruthiganesh.github.io/hackathons#proof-of-possible-2026-splitshield/Duplicate images leaking between train and test make model accuracy look better than it is. SplitShield finds the leakage, shows the evidence, repairs the split, and measures the difference.Sat, 22 Aug 2026 00:00:00 GMTHackathonhttps://parishruthiganesh.github.io/hackathons#cockroachdb-aws-agentic-memory/https://parishruthiganesh.github.io/hackathons#cockroachdb-aws-agentic-memory/An incident-response agent that remembers consequences.Mon, 17 Aug 2026 00:00:00 GMTHackathonhttps://parishruthiganesh.github.io/hackathons#build-with-gemini-xprize/https://parishruthiganesh.github.io/hackathons#build-with-gemini-xprize/An AI trading copilot for India's options traders.Mon, 17 Aug 2026 00:00:00 GMTHackathonhttps://parishruthiganesh.github.io/publications/intent-classification-41-models/https://parishruthiganesh.github.io/publications/intent-classification-41-models/A systematic zero-shot evaluation of 41 open-weight language models spanning 15 families and the 135M–9B parameter range, across eight English single-label intent-classification datasets covering standard benchmarks, a large-scale voice-assistant corpus, and production-derived e-commerce datasets. Beyond exact-match accuracy the study analyses confidence calibration, robustness to realistic input perturbations, statistical reliability of model rankings, deployment efficiency, and benchmark saturation. Results show that instruction-tuned 3B models can outperform several evaluated 7B base models, that differences among leading models on MASSIVE are statistically indistinguishable under pairwise McNemar tests, and that widely used benchmarks such as SNIPS have become saturated and no longer meaningfully discriminate among current open-weight models.Thu, 01 Jan 2026 00:00:00 GMTPublicationlanguage-model-evaluationhttps://parishruthiganesh.github.io/publications/interaction-representations-early-violence/https://parishruthiganesh.github.io/publications/interaction-representations-early-violence/Articulated human pose provides detailed body-configuration information beyond coarse spatial relationships, but whether this detail yields greater discriminative information when the downstream pipeline is held fixed remains unclear. We examine this through early violence detection. Holding the tracker, temporal head, supervision, folds, and evaluation fixed, we compare five interaction representations spanning coarse bounding-box geometry, a matched handcrafted pose analogue, enriched pose descriptors, and a matched-capacity encoder learned from raw joints, under video-level evaluation with cluster-bootstrap intervals. No pose-based representation outperforms coarse geometry, though with fifteen anomalous videos this subset cannot rule out small effects. Extending the pipeline to frozen visual encoders, and repeating the comparison on XD-Violence (137 anomalous videos, nine times our UCF- Crime sample), person-crop appearance and whole-frame context both exceed geometry by a wide margin, yet context matches appearance on UCF- Crime and exceeds it on the larger split: cropping to the interacting people yields no advantage over encoding the whole frame. This prompts a direct test of what the benchmark measures. Scoring anomalous videos using only frames preceding the annotated onset, under a control removing sequence length as a cue, retains 39-91% of above-chance separation on both benchmarks, including for seven hand-designed geometric channels. Inspection of the tightest pre-onset windows identifies concrete provenance artifacts: editorial title cards and platform watermarks absent from the surveillance footage supplying the normal class. Video-level AUC here is thus a composite of event evidence and pre-event source cues, a shared source of discrimination that can obscure differences between representations. The diagnostic requires only annotations these benchmarks already ship.Thu, 01 Jan 2026 00:00:00 GMTPublicationvideo-understanding-safetyhttps://parishruthiganesh.github.io/projects/intent-classification-benchmark/https://parishruthiganesh.github.io/projects/intent-classification-benchmark/A systematic zero-shot evaluation of 41 open-weight language models across eight intent-classification datasets.Thu, 01 Jan 2026 00:00:00 GMTProjectlanguage-model-evaluationhttps://parishruthiganesh.github.io/projects/early-violence-detection/https://parishruthiganesh.github.io/projects/early-violence-detection/A controlled ablation isolating how much explicit interaction modelling contributes to early detection in video.Thu, 01 Jan 2026 00:00:00 GMTProjectvideo-understanding-safetyhttps://parishruthiganesh.github.io/projects/sentinel-memory/https://parishruthiganesh.github.io/projects/sentinel-memory/An incident-response agent whose memory is a transactional database rather than a context window — it remembers consequences, not conversations.Thu, 01 Jan 2026 00:00:00 GMTProjectagentic-memory-systemshttps://parishruthiganesh.github.io/projects/astranova/https://parishruthiganesh.github.io/projects/astranova/An AI trading copilot for India's options traders — live market data, AI-scored signals, broker execution and a Gemini-powered assistant.Thu, 01 Jan 2026 00:00:00 GMTProjectapplied-ai-data-systemshttps://parishruthiganesh.github.io/projects/ausme-faculty-rag/https://parishruthiganesh.github.io/projects/ausme-faculty-rag/Retrieval-augmented search and summarisation over roughly 25,000 research papers, built to surface faculty expertise and collaboration opportunities.Thu, 01 Jan 2026 00:00:00 GMTProjectretrieval-augmented-generationhttps://parishruthiganesh.github.io/projects/dema-platform/https://parishruthiganesh.github.io/projects/dema-platform/A full-stack desktop application for engineering data lifecycle management, implementing the CIM4DE conceptual information model end to end.Thu, 01 Jan 2026 00:00:00 GMTProjectapplied-ai-data-systemshttps://parishruthiganesh.github.io/projects/rocket-simulation/https://parishruthiganesh.github.io/projects/rocket-simulation/A Unity simulation letting students change rocket parameters and watch the resulting flight behaviour in real time, across PC and tablet.Thu, 01 Jan 2026 00:00:00 GMTProjecthttps://parishruthiganesh.github.io/projects/bert-sentiment-analysis/https://parishruthiganesh.github.io/projects/bert-sentiment-analysis/Fine-tuning bert-base-uncased on the SMILE annotation dataset, as an end-to-end study of transformer transfer learning.Wed, 01 Jan 2025 00:00:00 GMTProjectlanguage-model-evaluationhttps://parishruthiganesh.github.io/projects/circular-trade-fraud-detection/https://parishruthiganesh.github.io/projects/circular-trade-fraud-detection/Detecting circular trading rings in iron-dealer invoice data using cycle detection, Node2Vec embeddings and density-based clustering.Mon, 01 Jan 2024 00:00:00 GMTProjectapplied-ai-data-systems