Hardik Jain

Indian Institute of Technology Jodhpur · India

Unclaimed · AI-structured profileOpenAlex

Early-career ML researcher at IIT Jodhpur in physics-informed neural networks and efficient 3D point cloud models

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-16) 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—

Output cannot be scored yet: the OpenAlex record shows an h-index of 0 and only 5 works indexed, all dated 2026. This axis would measure cumulative publication volume and yearly productivity once a fuller record exists.

Impact—

Impact is not yet scoreable: h-index and i10-index both stand at 0 with no citation counts available for the recent arXiv preprints. This axis would track citation accumulation and downstream uptake as the 2026 papers age.

Focus0/100

The focus score is 0, reflecting a record that is still dispersed. OpenAlex distributes the indexed works across five distinct topics — Robotics and Sensor-Based Localization (2), 3D Surveying and Cultural Heritage (2), plus three single-count neural-network and vision topics — while the reviewed output splits evenly (0.5/0.5) between scientific machine learning and efficient deep learning for 3D point clouds. With only two reviewed papers, the split is real but the picture remains thin.

Collaboration32/100

Collaboration scores 32 on a 0–100 scale, consistent with a small early-stage network: the OpenAlex record lists 4 distinct coauthors over the recent window. The recurring names are Anchit Mulye, Rhythm Baghel, Sujay Kumar Ingle, and Lakshay Chawla. On both cataloged 2026 preprints Jain sits in the last author position without corresponding-author status, suggesting a supervisory or senior-seat role within compact IIT Jodhpur teams rather than wide multi-institution networking.

Activity52/100

Activity scores 52. All indexed output is concentrated in a single year: countsByYear shows 5 works in 2026, and recent3yWorks is likewise 5, meaning no indexed publications appear in earlier years of the window. Within that burst, two arXiv preprints are cataloged — LPINNs and H3DNAS — spanning scientific machine learning and point-cloud efficiency. The score reflects genuine current momentum offset by a very short observable publication history.

Reproducibility55/100

Reproducibility scores 55, the mean over the 2 papers Evolve has reviewed (reproMean 55, reproPapers 2), both rated medium readiness. LPINNs presents clean paired-seed evidence on three toy benchmarks, though configuration choices were made post hoc and only a vanilla baseline is compared. H3DNAS reports CPU-only speedups on a single dataset with no variance measures, and its 'provable ceiling' claim is self-contradicted — solid artifacts, uneven statistical hygiene.

Research summary

Hardik Jain is an early-stage machine learning researcher at the Indian Institute of Technology Jodhpur, India, whose indexed work splits evenly (0.5/0.5) between two lines: scientific machine learning for neural PDE solvers and efficient deep learning with neural architecture search for 3D point clouds. Evolve's catalog includes two 2026 arXiv preprints, both with Jain in the last author position and neither as corresponding author. LPINNs introduces first-layer gated localization for physics-informed neural networks; the review found error reductions of one to two orders of magnitude across three toy benchmarks with clean paired-seed evidence, while noting post-hoc configuration choices and a vanilla-only baseline. H3DNAS is a hardware-aware pipeline that compresses 3D point cloud models directly from ONNX graphs; reviewers flagged a self-contradicted 'provable ceiling' claim, CPU-only speedups, and single-dataset, no-variance results. OpenAlex indexes 5 works, all dated 2026, with 4 distinct coauthors and h-index and i10-index both at 0, so output and impact remain unscored. Reproducibility averages 55 across the two reviewed papers, each rated medium readiness. OpenAlex topic labels on the record (robotics, surveying, cultural heritage) diverge from the reviewed fields and may reflect name-resolution noise.

Collaboration pattern

Collaboration is compact and local: the OpenAlex record lists 4 distinct coauthors (score 32), and the recurring circle is Anchit Mulye, Rhythm Baghel, Sujay Kumar Ingle, and Lakshay Chawla. Both cataloged 2026 preprints place Jain in the last author position without corresponding status, consistent with a senior or supervisory seat on small IIT Jodhpur teams. No multi-institution or international co-authorship is visible in the indexed window.

Research keywords

physics-informed neural networksfirst-layer gated localizationreceptive fieldsradial basis functionsdomain decomposition alternativeslong-domain PDEsneural architecture searchstructured pruningONNX-native compression3D point cloud processingedge deployment

Highlights

First-layer gating cuts PINN error by 1–2 orders of magnitude
LPINNs adds gated localization to the first layer of physics-informed neural networks, drawing on receptive-field and radial-basis ideas as an alternative to domain decomposition. On three toy benchmarks it reduces error by one to two orders of magnitude with clean paired-seed evidence — though configurations were chosen post hoc and only a vanilla baseline was tested.
Compressing 3D point cloud models straight from ONNX
H3DNAS couples structured pruning with hardware-aware neural architecture search over ONNX graphs, targeting edge deployment. Reported speedups are CPU-only, results come from a single dataset with no variance measures, and the 'provable ceiling' claim is self-contradicted — a useful engineering direction whose strongest claims remain unproven.
A dual-track 2026 output burst
OpenAlex indexes 5 works in 2026 — the researcher's entire recent record — split evenly between neural PDE solvers and 3D point cloud efficiency. Both cataloged papers emerge from the same four-coauthor circle at IIT Jodhpur, with Jain in the last author position, marking an early burst of senior-seat contributions across two fields.

Timeline

2026Released the LPINNs arXiv preprint, introducing first-layer gated localization for physics-informed neural networks, in the last author position.
2026Released the H3DNAS arXiv preprint on hardware-aware, ONNX-native 3D point cloud model compression, again in the last author position.
2026OpenAlex records 5 works in this single year — the only indexed output in the window — with 4 distinct coauthors.
2026Both preprints enter the Evolve catalog and are rated medium reproducibility readiness (mean 55).

Papers in the Evolve catalog

Frequent co-authors

Anchit Mulye×1 Rhythm Baghel×1 Sujay Kumar Ingle×1 Lakshay Chawla×1

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.

Bibliometrics come from OpenAlex only; no ORCID is linked, so 'Hardik Jain' is name-resolved from public bibliographic data and the record may mix identically named authors.
OpenAlex topic labels (robotics, surveying, cultural heritage) diverge from the reviewed paper fields and may reflect indexing or disambiguation noise.
Both cataloged papers are 2026 arXiv preprints with h-index and i10-index at 0; citation impact cannot yet be assessed.
Field distribution (0.5/0.5) and the reproducibility score derive from only 2 Evolve-reviewed papers; other indexed works are unreviewed.