About Me
Data Scientist on Meesho's ranking team, building large-scale multi-task deep-learning rankers (FT-Transformer, DCN-v2, Mixture-of-Experts) over 200+ feature spaces and 8-GPU distributed training. I ship models through the full loop — feature engineering, offline evaluation, A/B testing, and root-cause analysis — with measured platform-level impact. Before this, I did research in computational biology at the Indian Institute of Science, NTU Singapore, and the Centre for Brain Research, resulting in two peer-reviewed publications.
Contact Details
Email: ksrmanikumar@gmail.com
Phone: +91 94802 25156
LinkedIn: linkedin.com/in/manikumar-ksr
GitHub: github.com/Manikumarksr
Location: Bengaluru, India
Education
Indian Institute of Science (IISc)
Bangalore, India · 2020 – 2025
BS–MS (Research) — integrated 5-year Bachelor + Master's research degree · CGPA: 7.7/10
Master's thesis: The Vein Visualizer — Segmenting Hand Vein Images Using Deep Learning, Supervisor: Prof. Hardik J. Pandya, DESE
Bachelor's thesis: GPT-based Gene Function Information Retrieval from Scientific Literature, School of Biological Sciences, NTU Singapore
Key coursework: Probability & Statistics, Machine Learning for Data Science, Linear Algebra, Deep Learning, Generative AI, Computational Epidemiology
Blooms Pre-University College
Bagepalli, India · 2018 – 2020
Pre-University Education — 96.7% · Physics, Chemistry, Mathematics, Biology, English
Experience
Meesho — Data Scientist, Personalisation & Ranking
May 2025 – Present · Bengaluru, India
- Diagnosed a structural gap in the product-detail-page ranker: it was trained on data where two-thirds of orders were cash-on-delivery, systematically under-ranking prepaid-viable inventory even for users who almost always pay online.
- Designed a four-tier user-cohort taxonomy and a prepaid-propensity signal, then invented cohort-based feature splitting inside a single unified FT-Transformer multi-task ranker — avoiding the serving cost of per-cohort models.
- Shipped to a live A/B (since Feb 2026): +1.05% orders per visitor on the product page, +0.48% platform-wide, with the best-performing cohort seeing +2.79%.
- Reverse-engineered and documented the company-wide offline feature store (~140 Delta tables, 11 entity types) — now a shared reference across recommendations, search, fraud and logistics teams.
Fujitsu Research of India Pvt. Ltd. — AI/ML Research Intern
May 2024 – Aug 2024 · Bengaluru, India
- Surveyed the state of the art in explainability for time-series forecasting and implemented explainable forecasting models.
- Built and deployed real-time Streamlit dashboards exposing model explanations (SHAP, LIME) to non-technical stakeholders.
Nanyang Technological University — Research Intern, Mutwil Lab
Aug 2023 – Dec 2023 · Singapore · NTU–India Connect Research Fellow
- Led an LLM pipeline that mined 84,427 publication abstracts to construct a yeast knowledge graph of 3.4M+ entities, live at yeast.connectome.tools.
- Published as first author in the Journal of Molecular Biology (2025).
Centre for Brain Research (CBRAIN), IISc — Research Intern
May 2023 – Jul 2023 · Bangalore, India
- Analysed gene-expression data from smokers and non-smokers to discover novel expression quantitative trait loci (eQTLs), using QTLtools and BCFtools for variant calling and association testing.
Scholarships & Awards
- KVPY Fellow (Kishore Vaigyanik Protsahan Yojana) — national science-talent fellowship, Govt. of India, top ~1% of applicants nationally · 2020 – 2025
- NTU–India Connect Research Fellow — competitive international research fellowship, NTU Singapore · Aug – Dec 2023
- Winner, TATA Building India Essay Competition — national-level · 2017
- District-level Winner, Scientific Model on Natural Resource Management · 2016