Azade Farshad

Research profile

Research summary

Azade Farshad is a researcher at the Munich Center for Machine Learning whose work spans two fields in near-equal measure: machine learning for healthcare with a precision-nutrition emphasis, and computational biology / medical AI. The OpenAlex-indexed record (identity: openalex) shows h-index 9 and i10-index 8, with 27 works in the most recent three-year window and a steep publication ramp — from 3 indexed works in 2020 to 10 in 2022 and a peak of 14 in 2025. Topical breadth is characteristic: generative adversarial networks and image synthesis (23 works), advanced neural network applications (22), multimodal machine learning (18), domain adaptation and few-shot learning (16), and medical image segmentation (15). Evolve has reviewed two 2026 arXiv preprints. In Predicting Postprandial Glycemic Response from Meal Images, Clinical Variables, and Gut Microbiome Information, Farshad is last author on a joint image–macronutrient model with attention-FiLM fusion reaching R=0.74 on CGMacros — promising, though a 7-subject test set and thin baselines cap the evidence. In HPOQuest, a middle-author contribution, a training-free LLM agent performs sequential Human Phenotype Ontology questioning for rare-disease diagnosis — strong on a small MME cohort, mixed elsewhere, with no significance tests reported. Reproducibility averages 55 across the two reviewed papers (both medium readiness), and the collaboration network is compact: 7 distinct coauthors with repeat ties to Kamilia Zaripova and Nassir Navab.

Research keywords

Highlights

• Postprandial glucose prediction from meal images, clinical variables, and gut microbiome

  As last author, contributed to a joint image–macronutrient model with attention-FiLM fusion on the CGMacros dataset, reaching R=0.74 for postprandial glycemic response. Evolve judges the approach promising but capped by a 7-subject test set and thin baselines.

• HPOQuest: training-free rare-disease diagnosis via active phenotype acquisition

  Middle-author contribution to an LLM-based diagnostic agent that sequentially acquires Human Phenotype Ontology terms to improve sparse rare-disease diagnosis. Results are strong on the small MME cohort, mixed elsewhere, with no significance tests reported.

• Sustained publication ramp to a 2025 peak

  OpenAlex-indexed output grew from 3 works (2020) to 6 (2021) and 10 (2022), held at 9 and 8 through 2023–2024, and peaked at 14 indexed works in 2025, with 27 works over the most recent three-year window.

Timeline

1. 2014Earliest indexed work appears in the OpenAlex record (1 publication); activity stays sparse for years, with only 1 work indexed again in 2019.
2. 2020Output begins to build: 3 indexed works in 2020, doubling to 6 in 2021.
3. 202210 indexed works — the strongest year of the early ramp — followed by sustained output of 9 works in 2023 and 8 in 2024.
4. 2025Peak year with 14 indexed works; the record now counts 27 works over the most recent three-year window, with h-index 9 and i10-index 8.
5. 2026Two arXiv preprints enter the Evolve catalog — the postprandial glycemic response model (last author) and HPOQuest for rare-disease diagnosis (middle author); 5 works are indexed so far this year.

Papers in the Evolve catalog

• 2026-09-21

  Predicting Postprandial Glycemic Response from Meal Images, Clinical Variables, and Gut Microbiome Information

• 2026-09-17

  HPOQuest: A Rare-Disease Diagnostic Agent Using Active Phenotype Acquisition

Frequent co-authors