Apple ML researcher spanning privacy-preserving federated learning for speech and applied marketing analytics
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
5
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—
Not scored: the snapshot lacks complete, deduplicated work-level metadata, so total publication volume and its year-over-year consistency cannot be computed reliably. This axis would measure cumulative output and publication cadence across the career.
Impact—
No impact score is available: citation-per-paper coverage in the snapshot is insufficient for a reliable measure. This axis would capture citations, citation trajectory, and influence on later work; the reported h-index of 5 alone cannot substitute for it.
Focus0/100
Focus scores 0 because the profile splits evenly across two distant fields — machine learning for speech processing/federated learning and applied ML for marketing analytics — at 0.5 weight each. Topic counts reinforce breadth over a single line: privacy-preserving technologies 11, stochastic gradient optimization 6, domain adaptation and few-shot learning 4, multi-agent control 3, IoT/edge computing 3.
Collaboration48/100
Collaboration scores 48 on 10 distinct coauthors. The ten most frequent collaborators include Wonho Bae, Zakaria Aldeneh, Tatiana Likhomanenko, Jan "Honza" Silovsky, Tianqi Wang, Martin Pelikan, Christopher G. Brinton, and Jan P. Allebach. On both reviewed 2026 papers Azam sits in a supporting position — last author on the federated ASR preprint, middle author on the B2B conversion paper — and is corresponding author on neither.
Activity32/100
Activity scores 32. By-year counts run 2 (2018), 1 (2019), a steady 3 per year through 2020-2022, 1 in 2023, then 2 in 2026; recent3yWorks is 2. Recent output sits well below the earlier three-year plateau, a cadence consistent with industry research where publication rate is uneven.
Reproducibility38/100
Reproducibility averages 38 across the 2 papers Evolve reviewed. The federated ASR paper earns medium readiness — ablations are convincing but evidence is single-run and simulation-only — while the B2B CatBoost pipeline reports 91% accuracy on proprietary HP data with weak evaluation and no external validation, rated low. Neither paper has been independently validated.
Research summary
Sheikh Shams Azam is a researcher at Apple (United States) whose public bibliographic record spans two evenly split halves: machine learning for speech processing and federated learning (field weight 0.5) and applied machine learning for marketing analytics (0.5). The OpenAlex-resolved snapshot records 15 works across 2018-2026, led by the topic Privacy-Preserving Technologies in Data (11 works), followed by Stochastic Gradient Optimization Techniques (6), Domain Adaptation and Few-Shot Learning (4), Distributed Control Multi-Agent Systems (3), and IoT and Edge/Fog Computing (3). The snapshot reports an h-index of 5 and an i10-index of 4. Evolve has reviewed two 2026 arXiv preprints: a practical recipe for semi-supervised federated ASR built on online pseudo-labels with server update stabilization (last author), judged to have convincing ablations but single-run, simulation-only evidence; and a B2B customer conversion methodology combining document representation, graph theory, and CatBoost (middle author), credited with solid engineering but weak evaluation, no external validation, and a claimed 91% accuracy on proprietary HP data. Recent activity is thin — recent3yWorks is 2, versus peaks of 3 works per year in 2020-2022 — and the researcher collaborates with 10 distinct coauthors, typically in supporting author positions.
Collaboration pattern
The snapshot lists 10 distinct coauthors; the ten most frequent include Wonho Bae, Zakaria Aldeneh, Tatiana Likhomanenko, Jan "Honza" Silovsky, Tianqi Wang, Martin Pelikan, Wan Eih Huang, Anton Wiranata, Christopher G. Brinton, and Jan P. Allebach. On both reviewed 2026 papers Azam takes a supporting position — last author on the federated ASR preprint and middle author on the B2B conversion paper — with neither role as corresponding author, a pattern typical of industry team research.
Research keywords
Highlights
Server update stabilization for semi-supervised federated ASR
Graph- and CatBoost-driven B2B conversion prediction
Privacy-preserving machine learning as the dominant topic
Privacy-Preserving Technologies in Data is the most frequent topic in the bibliographic topic profile with 11 works, ahead of stochastic gradient optimization techniques (6) and domain adaptation and few-shot learning (4), anchoring the federated learning line of work.
Timeline
- 2018Earliest year in the bibliometric snapshot, with 2 works recorded.
- 2020-2022Publication plateau: 3 works recorded in each of 2020, 2021, and 2022 — the most productive stretch in the record.
- 2023Only 1 work recorded, the lowest point since 2019.
- 2026Two arXiv preprints enter the Evolve catalog: the semi-supervised federated ASR recipe (last author) and the B2B CatBoost conversion methodology (middle author).
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.
- Bibliometrics derive from an OpenAlex identity record (ORCID 0000-0001-7678-5092); the profile is name-resolved from public bibliographic data, and output/impact could not be scored from it.
- Only two papers — both 2026 arXiv preprints — have been reviewed by Evolve, so quality and reproducibility judgments are preliminary.
- The h-index (5) and i10-index (4) are quoted from the snapshot as reported, without independent verification of the underlying citation list.
- Field and topic labels are algorithmic OpenAlex assignments, not self-authored research statements.