Industrial ML researcher at LinkedIn building GPU-powered retrieval and pre-ranking for large-scale recommender systems
2
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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—
Not yet scored: this profile has too few indexed publications for a reliable output count. The output axis would measure total publication volume and its distribution over venues and years.
Impact—
No citation metrics are available yet, so impact is not scored. This axis would measure citation counts, citing works, and the uptake of the researcher's retrieval and pre-ranking methods by others.
Focus100/100
Scores 100 for focus: the only field-classified paper in the catalog sits in recommender systems / information retrieval, and both 2026 papers — the LinkedIn Connected Content Retriever and Efficient GPU Retrieval for Semantic Search — target GPU-accelerated retrieval. With so few classified records, this concentration should be read as an early-stage signal rather than a long-term specialization.
Collaboration66/100
Scores 66 for collaboration, grounded in 27 distinct coauthors and 12 recurring collaborators in the record. The pattern is typical of industrial team research: large, role-specialized author groups at LinkedIn, with the researcher first-authoring one 2026 paper and contributing as a middle author on the LinkedIn pre-ranking systems paper.
Activity—
Not yet scored: publication timestamps are too sparse to compute a reliable activity rhythm. This axis would measure output cadence across years, such as papers per year and gaps between publications.
Reproducibility20/100
Scores 20 for reproducibility, computed from one assessed paper with a mean readiness of 20. Evolve's review of the Connected Content Retriever labels its reproducibility readiness low: the reported +2.5% lift in Feed time spent is credible as a production result, but training and baseline details are thin, consistent with that low label and making independent replication difficult.
Research summary
Dhritiman Das is an industrial machine-learning researcher affiliated with LinkedIn (United States), profiled here from public bibliographic records under ORCID 0000-0001-6627-0618. The Evolve catalog tracks two of his 2026 arXiv papers, both centered on GPU-accelerated retrieval. The first-authored 'Efficient GPU Retrieval for Semantic Search' is cataloged but not yet reviewed. The second, 'Connected Content Retriever: Dense Graph Edge Features Powering Pre-Ranking at LinkedIn', lists him as a middle author and describes LinkedIn's GPU pre-ranker, which joins dense edge features from the social graph at runtime; Evolve's review records a +2.5% lift in Feed time spent and calls it solid systems work while flagging thin training and baseline details, with reproducibility readiness rated low. The radar reflects this early, industry-shaped record: focus 100, since the only field-classified paper sits in recommender systems / information retrieval; collaboration 66, from 27 distinct coauthors with 12 recurring; and reproducibility 20, a mean over one assessed paper. Output, impact, and activity remain unscored because the underlying data is insufficient. Overall, this is an early-stage profile of a systems-oriented retrieval researcher whose visible work targets pre-ranking and semantic-search retrieval at industrial scale, within large specialized teams rather than as a solo line of research.
Collaboration pattern
The record shows 27 distinct coauthors and 12 recurring collaborators, including Chirag Bhanuprasad Mehta, Chujie Zheng, Ronak Kaoshik, Pratik Dixit, Vishal Shah, Yanbo Li, Jiahao Xu, Manika Agarwal, Chinmay Naik, Lingyu Zhang, Chetan Bhole, and Meng Zheng, consistent with large industrial author teams at LinkedIn. Das appears as first author on one 2026 paper and as a middle author on the LinkedIn systems paper, pointing to contribution within a team rather than solo work.
Research keywords
Highlights
GPU pre-ranking with runtime graph edge-feature joins at LinkedIn
Retrieval-as-ranking over the social graph
First-author work on efficient GPU retrieval for semantic search
Das is first author of 'Efficient GPU Retrieval for Semantic Search' (arXiv, 2026), a paper on GPU-efficient retrieval for semantic search. It is in the Evolve catalog but has not yet been reviewed, so no verdict, keywords, or reproducibility assessment exists for it.
Timeline
- 2026Co-authored (middle author) the arXiv paper 'Connected Content Retriever: Dense Graph Edge Features Powering Pre-Ranking at LinkedIn' on LinkedIn's industrial GPU pre-ranker.
- 2026First-authored the arXiv paper 'Efficient GPU Retrieval for Semantic Search' on GPU-efficient retrieval for semantic search.
- 2026Published two catalog-tracked papers within the same year, both in GPU-accelerated retrieval and pre-ranking, matching the concentrated focus the radar records (focus score 100).
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 scores are computed from OpenAlex-derived records plus Evolve's own paper reviews; coverage here is limited to two papers, both from 2026 on arXiv.
- Output, impact, and activity axes are unscored because citation counts, venue history, and a full publication list are not available in this snapshot.
- The profile is name-resolved from public bibliographic data (ORCID 0000-0001-6627-0618); affiliation is listed as 'LinkedIn (United States)' with no country or affiliation-history detail.
- 'Efficient GPU Retrieval for Semantic Search' has not been reviewed by Evolve, so its verdict, keywords, and reproducibility readiness are missing.