Md. Asif Uddin

    The record

    Published papers

    Papers and manuscripts, with a link wherever there is one to give. What a piece of work argues lives on its plate; this page is only the citation.

    Three parts. The first is mine. The second is the reviews under way — method rather than manuscript, since nothing is registered until it is. The third is everyone else's, in the order they make sense.

    The recordMine

    1. Published2026

      HierarchiRetina: A Multi-Stage Deep Learning Pipeline for Diabetic Retinopathy Severity Assessment

      Md. Asif Uddin

      BSc thesis, BRAC University

      Interpretable diabetic retinopathy grading from retinal photographs. Deposited in the BRAC University institutional repository.

    2. Under review2026

      Rank Radii Transfer as Quantiles: Taxonomic Placement and Sub-Taxonomy Construction for Unseen Species

      Md. Asif Uddin, Niloy Farhan

      Placing unseen species inside the taxonomy rather than in flat clusters.

    3. In preparation2026

      Automated Renal Reporting

      Md. Asif Uddin, Niloy Farhan

      Measure first, write second — the kidney section of a CT report, generated from computed values a verifier will not let the model drop.

    ReviewsUnder way

    1. ProtocolCausal inference · single-cell genomics · Alzheimer's microglia

      How perturbation-prediction methods are validated

      What evidence do published methods offer that their counterfactual predictions hold — and against what was the comparison made.

      Reporting
      PRISMA-ScR
      Registry
      OSF Registries
      Instrument
      The Validation Ladder — five tiers, from none, through known biology and held-out observational data, to prospective experiment. Only the comparison target sets the tier; interventional data used as input does not.
      Scale
      32 records piloted, 16 included
      Supports
      HCGT-PG

    2. ProtocolMedical imaging · segmentation · report generation

      Automated renal and urinary tract assessment on CT

      Which structures and tasks the published literature automates, on which datasets, and how the results are evaluated.

      Reporting
      PRISMA-ScR, JBI methodology
      Registry
      OSF Registries
      Instrument
      A fixed anatomical vocabulary, ticked against every record, so a structure nobody has addressed is recorded as absent rather than inferred from silence.
      Scale
      Five databases — PubMed, Embase, Scopus, IEEE Xplore, arXiv — with about twenty-five charted fields per paper.

    LibraryReadable by the machine

    Papers I have added for ResearchLens to read. It indexes them rather than learning them, so anything it says from one arrives with a citation to the passage it came from.

    1. An introduction to artificial neural networks

      Coryn A.L. Bailer-Jones, Ranjan Gupta, Harinder P. Singh · 2001

    2. A Clinically Informed Two-Stage Framework for Renal CT Report Generation

      Renjie Liang, MSca, Zhengkang Fan, MSca, Jinqian Pan, MSca, Chenkun Sun, MSca, Bruce Daniel Steinberg, MDb, Russell Terry, MDb, Jie Xu, PhDa, · 2025

    3. SUPERINTELLIGENCE: Paths, Dangers, Strategies

      NICK BOSTROM · 2014

      We begin by looking back. History, at the largest scale, seems to exhibit a sequence of distinct growth modes, each much more rapid than its predecessor. This pattern has been taken to suggest that another (even faster) growth mode might be possible. However, we do not place much weight on this observation—this is not a book about “technological acceleration” or “exponential growth” or the miscellaneous notions sometimes gathered under the rubric of “the singularity. ” Next, we review the history of artificial intelligence. We then survey the field’s current capabilities. Finally, we glance at some recent expert opinion surveys, and contemplate our ignorance about the timeline of future advances.

    4. Introduction to Algorithms

      Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest, Clifford Stein · Fourth Edition [2022]

      When you design and analyze algorithms, you need to be able to describe how they operate and how to design them. You also need some mathematical tools to show that your algorithms do the right thing and do it efficiently. This part will get you started. Later parts of this book will build upon this base.

    5. Algorithms to Live By

      Brian Christian & Tom Griffiths · 2016

      It takes concepts like sorting, optimal stopping, exploration vs. exploitation, Bayesian reasoning, scheduling and game theory and connects them to everyday decisions.

    LectionesEveryone else's, in order

    A reading course, published in parts of three to five papers. Every paper appears exactly once, so finishing a part means something. 12 parts so far, 54 papers, each with what it claims and why it is on the list.

    1. FoundationsWhat a deep network is, and the four papers that settled it.4
    2. The TransformerOne architecture, and the three papers that showed what it could carry.4
    3. AlignmentTurning a model that predicts text into one that answers you.5
    4. Generating before the language modelFour ways to learn a distribution you can sample from.4
    5. Vision after the TransformerWhat happened to computer vision once attention arrived.5
    6. Language and vision in one modelHow a model comes to see and talk about the same thing.5
    7. Position, memory and the cost of attentionAttention is quadratic and the cache does not fit. Five papers about that.5
    8. Fine-tuning without the computeThree papers, five years, and the reason you can tune a large model on one GPU.3
    9. Sparsity and scaleWhy a model can have 671 billion parameters and use 37 billion of them.5
    10. ReasoningFour papers in which nothing about the model changes and the answers get better.4
    11. RetrievalGiving a model access to things it was not trained on.5
    12. AgentsWhere the model stops answering and starts doing.5

    The whole course