Time series ML researcher at Huawei Sweden: forecasting, anomaly detection, and open time-series foundation models
3
papers here
3
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—
No aggregate output metrics — total publication count, venue mix, or yearly volume — are available in the current snapshot, so this axis is unscored. It would measure the scale and consistency of the publication record beyond the three reviewed preprints.
Impact—
Citation counts and h-index are not yet available for this profile, so impact cannot be scored. This axis would measure how often and how widely the reviewed preprints and any earlier work are cited by the research community.
Focus0/100
The focus score is 0, reflecting a field distribution split exactly into thirds across time series forecasting, time series anomaly detection, and time series foundation models — breadth rather than a single deep lane. All three lines sit inside the time-series ML umbrella, so this reads as one coherent program covering detection, forecasting, and pretrained models rather than scattered interests.
Collaboration51/100
The collaboration score is 51 (mid-range), computed from 12 distinct coauthors across the three reviewed papers. The frequent-coauthor list includes Ambroise Odonnat, Youssef Attia El Hili, Bahaeddine Abdessalem, Shifeng Xie, and Lujia Pan, consistent with team-based industrial research. Tiomoko appears as a middle or last author on all three papers without corresponding-author designation, suggesting a contributing rather than leading-corresponding role.
Activity—
Year-by-year publication data is insufficient in the current snapshot, so publishing cadence cannot be scored. This axis would measure how consistently the researcher produces work over time and whether the 2026 cluster of three preprints marks an acceleration or a one-off burst.
Reproducibility55/100
Reproducibility averages 55 across three papers (reproPapers: 3), each rated medium readiness. The signal is differentiated: Tabby is singled out by Evolve as unusually transparent and well-ablated, while FlowTSFM has pending ablations and a confounded CosMean baseline, leaving its evidence preliminary. The anomaly-detection paper's TSB-AD-U pipeline is judged convincing, but its ten-feature MAD design remains medium to reproduce.
Research summary
Malik Tiomoko is a machine learning researcher at Huawei Technologies (Sweden) whose reviewed work concentrates on the time-series stack: forecasting, anomaly detection, and time-series foundation models. The Evolve catalog holds three 2026 arXiv preprints, and the field distribution splits evenly — one third forecasting, one third anomaly detection, one third foundation models — indicating a deliberate sweep across the time-series landscape rather than a single narrow specialty. On anomaly detection, the LLM-generated feature pools paper couples a compact ten-feature median/MAD pipeline with LLM-written feature code, reaching 0.588 on TSB-AD-U and matching the best pretrained detectors; Evolve judges the claim convincing. On foundation models, Tabby (145M parameters) publishes an unusually transparent, well-ablated open pretraining recipe built on structural causal models and synthetic data, strong in-distribution on GIFT-Eval though mid-table on out-of-distribution TIME. FlowTSFM (38.8M) reframes encoder depth as recurrent quantile transport with competitive benchmark results, though ablations remain pending. Across the three papers Tiomoko works with 12 distinct coauthors and appears as a middle or last author without corresponding-author designation, consistent with a contributing role inside a Huawei industrial-research group. Reproducibility averages 55 across the three papers, all rated medium readiness.
Collaboration pattern
Tiomoko's reviewed work is thoroughly collaborative: 12 distinct coauthors across three 2026 preprints, with a stable Huawei-anchored circle including Ambroise Odonnat, Youssef Attia El Hili, Bahaeddine Abdessalem, Shifeng Xie, Lujia Pan, Keli Zhang, and Jianfeng Zhang. The pattern — middle or last authorship, never corresponding — fits an experienced contributor inside an industrial time-series group fielding concurrent papers on detection, forecasting, and foundation models.
Research keywords
Highlights
Open-recipe Tabby foundation model (145M) with unusually transparent pretraining
LLM-generated feature pools lift a compact MAD detector to parity on TSB-AD-U
FlowTSFM recasts encoder depth as recurrent quantile transport (38.8M)
Timeline
- 2026Publishes the LLM-generated feature pools approach for time series anomaly detection on arXiv (middle author), reaching 0.588 on TSB-AD-U.
- 2026Releases Tabby, an open pretraining recipe for a 145M time series foundation model (last author).
- 2026Proposes FlowTSFM, turning encoder depth into quantile transport with competitive benchmark results (last 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.
- Bibliometric scores are computed from an OpenAlex-derived snapshot only; citation counts, h-index, and aggregate output metrics were unavailable, so output, impact, and activity axes are unscored.
- The evidence base is three 2026 arXiv preprints reviewed by Evolve; peer-reviewed venues and any earlier or unpublished work fall outside this catalog, making this an early-stage profile.
- No ORCID or disambiguation data was provided; this profile is name-resolved from public bibliographic data and may merge or split identically named authors.
- All three papers are preprints; Evolve verdicts are independent review assessments, not community citation validation.