UT Arlington ML researcher in synthetic data: LLM-generated time series and distilled video-language models
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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 data (publication counts, venue spread, yearly volume) is not available in this snapshot, so this axis is unscored; it would measure total scholarly production and its consistency over time.
Impact—
Citation and influence metrics are not yet available for this researcher, so impact is unscored; it would measure citation uptake, citing works and the downstream influence of the reviewed papers.
Focus0/100
This axis scores 0. The recorded field distribution splits exactly 50/50 between two distinct directions: machine learning for synthetic time-series data generation in manufacturing, and machine learning for multimodal video-language models. With only two papers, one in each area, no dominant research line has emerged yet; the score reflects an even split rather than an established concentration, which may consolidate as the publication record grows.
Collaboration41/100
This axis scores 41. The bibliometric snapshot records seven distinct coauthors — including Mantek Singh, Prateek Karnal, Gagan Ganapathy, Vineet Shah, Ridam Arora, Siddharth Raina and Jasmin Jarsania — across two papers. Both cataloged papers place Challagundla in a middle author position without corresponding-author status, indicating work within team-led projects rather than independent leadership; the moderate score reflects a real but still small collaboration footprint.
Activity—
Recent-activity signals (publication cadence, preprint recency, ongoing projects) are insufficient in this snapshot, so this axis is unscored; it would measure how actively the researcher publishes over time.
Reproducibility38/100
This axis scores 38, a mean computed over two reviewed papers. The video-language reasoning-distillation paper is rated medium: results appear promising but come from a single run with no variance reported, and the key chain-of-thought-order ablation is unquantified. The LLM-based manufacturing time-series paper is rated low, with no significance tests, weak baselines and private data. Both verdicts point to a need for stronger evaluation rigor.
Research summary
Jeshwanth Challagundla is a machine-learning researcher at The University of Texas at Arlington whose reviewed work splits evenly between two directions: LLM-based synthetic time-series generation for manufacturing and multimodal video-language models. The manufacturing paper, 'LLMs as Master Forgers', asks whether LLMs with retrieval-augmented generation can synthesize industrial time series beating ARIMA and LSTM baselines; Evolve finds the results plausible but thin — no significance tests, weak baselines, private data — and rates reproducibility low. The video-language paper, 'Efficient Reasoning Distillation', reports a 2B-parameter model approaching its 4B teacher with roughly 900 chain-of-thought-distilled samples chosen by uncertainty and difficulty; Evolve calls it promising yet single-run, without variance reporting or a quantified CoT-order ablation, rating reproducibility medium. The bibliometric base is two 2026 arXiv papers, both in a middle author position with no corresponding-author role, seven distinct coauthors, and a reproducibility mean of 38 across both papers. The field distribution is a clean 50/50 split between the two areas, so the focus axis currently scores zero. Output, impact and activity axes lack sufficient data and remain unscored. This is an early-stage profile of a researcher whose unifying thread is synthetic data and efficient training, with the evidence base still accumulating.
Collaboration pattern
All cataloged work is collaborative: two 2026 arXiv papers with a middle author position and no corresponding-author role. The snapshot records seven distinct coauthors — Mantek Singh, Prateek Karnal, Gagan Ganapathy, Vineet Shah, Ridam Arora, Siddharth Raina and Jasmin Jarsania — who form the frequent-coauthor set. No senior-mentor or external-industry collaborators can be inferred from the available data, and citation-based collaboration analysis is not yet possible.
Research keywords
Highlights
Distilling video-language reasoning into a 2B model
LLMs as generators of manufacturing time series
A unifying thread: synthetic data for efficient ML
Across both reviewed papers, synthetic data is the connective tissue — synthetic chain-of-thought samples for distilling small video-language models, and LLM-generated time series for manufacturing — positioning this early-stage researcher around data-efficient machine learning rather than a single architecture or application.
Timeline
- 2026First cataloged arXiv release: 'LLMs as Master Forgers' applies LLMs with retrieval-augmented generation to synthetic manufacturing time series, with a middle author position and no corresponding-author role.
- 2026Second cataloged arXiv release: 'Efficient Reasoning Distillation' reports small video-language models distilled via synthetic chain-of-thought with difficulty-aware fine-tuning.
- 2026Both 2026 papers enter the Evolve catalog, establishing a 50/50 field split between manufacturing synthetic data and video-language modeling, with seven distinct coauthors recorded across the two papers.
Papers in the Evolve catalog
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 dimensions are computed from public bibliometric data only; the output, impact and activity axes lack sufficient data and are unscored.
- This profile rests on just two Evolve-reviewed 2026 arXiv papers; no peer-reviewed venue, citation-count or funding data is included.
- Reproducibility (mean 38) reflects Evolve's review of these two papers only, with readiness rated medium and low respectively.
- The profile is name-resolved from public bibliographic data (ORCID 0009-0005-5106-3743); country is not recorded.