Prakul Hiremath

Visvesvaraya Technological University (VTU), Belagavi, Karnataka, India

CGPA: 8.02/10

Studies conditions under which sequential detection can be represented by a scalar statistic without losing decision-relevant information.

Poster page · Sydney poster session: 10 December 2026, 10:00 AM–1:00 PM AEDT.

Studies how rollout latency and asynchronous data collection can introduce selection bias in reinforcement learning from human feedback.

Poster page · Sydney poster session: 8 December 2026, 5:00–8:00 PM AEDT.

Examines when scalar evaluation metrics fail to distinguish systems with meaningfully different behaviour.

OpenReview · Manuscript PDF

A post-hoc diagnostic for assessing whether frozen representations preserve task-relevant information under anticipated distribution shifts.

OpenReview · Manuscript PDF

arXiv:2606.11211

Studies alignment between uncertainty estimates and model sensitivity to input changes in high-risk prediction settings.

arXiv preprint · Code

Working with Dr. M. N. Birje on how model capabilities emerge during training and how detection timing depends on evaluation resolution. Studies the distinction between a capability first becoming detectable and becoming reliably usable across training checkpoints.

Presented “AI-Driven, Uncertainty-Aware Anomaly Detection in Multi-wavelength AstroSat Observations” at A Decade of AstroSat Observations: Science Outcomes and Future Prospects, organised by the U. R. Rao Satellite Centre (URSC), ISRO, Bengaluru, 30 January–1 February 2026.

Conference information

Contributed to BIOLOOP, an AI-enabled algal biorefinery concept for industrial CO₂ capture and biomass production, and to S2N, a seawater-based green-hydrogen initiative. BIOLOOP was included in an AI-in-energy case study for the India AI Impact Summit 2026.

Presented “Real-Time Monitoring of Bioaerosol Emissions in Hydrogen Reactors” at the 14th Asian Aerosol Conference (AAC 2025), Mumbai, December 2025.

Global AI Impact Commons case study · Poster story · Casebook announcement

Research on latent microstructure regimes and early signals in high-frequency limit order book data. Paper · Repository

Tool for detecting look-ahead bias and accidental use of future data in time-series machine-learning pipelines. PyPI · Repository · Project note

A C++ and Gymnasium environment for exploring reinforcement-learning approaches to latency optimisation in trading infrastructure. Repository · Project note

Multivariate temporal correlation engine for anomaly detection across related infrastructure and financial metrics. PyPI · Repository

Time-travel debugging and transformation-lineage tracking for Pandas and Polars workflows. PyPI · Repository

A time-aware metric designed to evaluate when a detector identifies an event alongside detection quality. Paper · PyPI · Colab

Algae-based approach to industrial CO₂ capture and biomass production. Contributed before exiting the initiative.

Seawater-based green-hydrogen initiative exploring pathways for hydrogen production.

Patent applications

• 202541014688 — Integrating Neuromorphic Computing and Reinforcement Learning in Hybrid AI for Autonomous Systems (2025; application).
• 202541014687 — Innovative CO₂ Emission Filters for Vehicles: Net-Zero Carbon Solutions with Advanced Adsorption and Smart Monitoring (2025; application).
• 202541014690 — Dynamic Adjustable Sleeve D&M for Enhanced Universal CO₂ Emission Filters (2025; application).
• 202541014689 — Hyper-Adaptive Manufacturing Using Multi-Agent AI for Autonomous Production Optimization (2025; application).
• 202541032942 — EMEFF: An Explainable Multimodal Economic Forecasting Framework for Real-Time, Transparent, and Adaptive Predictions (2025; application).
• 202641023750 A — A Spatial–Temporal Deep Learning Framework for Robust Human Pose Estimation in Crowded Visual Scenes (filed 27 February 2026; published 6 March 2026).

Registered designs

• 463354-001 — Modular Algae Pod Device for Sustainable CO₂ Capture (certificate issued 2 June 2026).
• 463353-001 — Dual-Chamber CO₂ Capture and Air Purification Device Using Algae on Biodegradable Film (certificate issued 25 March 2026).
• 463624-001 — Disposable Algae Bioreactor Device for CO₂ Capture Using Hydrogel and Bioplastics (certificate issued 25 March 2026).