Key Takeaways
- Modern AI excels at summarizing text but often fails to reliably convert unstructured content into accurate tables, a task critical for medical decision‑making.
- Naman Ahuja’s master’s research proposes a step‑wise AI pipeline—extract atomic facts, plan table structure, then incrementally fill the table—to improve traceability and reduce errors.
- The method mirrors human reasoning, allowing clinicians to reuse structured data from heterogeneous sources like PDFs or Wikipedia without repetitive manual extraction.
- By supporting “living data,” the system can update existing tables as new studies appear, preserving consistency while incorporating fresh evidence.
- Ahuja’s work earned him an IBM Infrastructure Master’s Fellowship Award, recognizing its real‑world impact and potential for trusted AI in health‑care settings.
- After graduating from ASU’s Ira A. Fulton Schools of Engineering, Ahuja will join Amazon in Seattle to continue building large‑scale systems that translate research into usable industry solutions.
The Problem with Current AI‑Generated Tables
Modern medicine depends on quickly scannable data—survival rates, drug side effects, trial outcomes—often presented in tables that clinicians can read in seconds. Yet producing those tables remains a bottleneck: researchers must manually pull facts from dense PDFs, journal articles, or web pages and organize them into a usable format. As Ahuja observes, “Large language models can read and summarize documents with ease. But when asked to extract precise information and organize it into something structured — like a table a doctor or analyst could rely on — they often struggle.” The resulting AI‑generated tables may look polished but frequently miss details, introduce inconsistencies, or hallucinate unsupported claims, jeopardizing the reliability clinicians need for high‑stakes decisions.
A Step‑wise Solution Inspired by Human Workflow
To close this gap, Ahuja’s master’s thesis rethinks how AI should approach table generation. Instead of demanding a single‑shot output, he breaks the task into three verifiable stages: first, the system extracts atomic facts from the source text; second, it devises a plan for how the final table should be organized; third, it incrementally populates the table, updating entries as new information emerges. This mirrors the way a human analyst would work—reading carefully, deciding which categories matter, then filling in the spreadsheet piece by piece. By decomposing the problem, the approach improves traceability and reduces the chance of error, a crucial advantage when lives depend on the accuracy of the data.
Why Traceability Matters in Health Care
In systematic reviews, clinicians sift through large volumes of research to extract key findings into evidence tables that guide treatment choices. Mistakes in this process can lead to flawed guidelines or inappropriate prescriptions. Ahuja emphasizes, “We can use these heterogeneous documents and convert them into a structured data form so that we can access that information more readily… That can help reduce repetitive data extraction and allow clinicians to focus more on interpreting results.” His method ensures each cell in the table can be traced back to its originating sentence, providing auditors and reviewers a clear lineage of information—a feature that single‑pass generation lacks.
Handling Evolving, “Living” Data
Medical knowledge is not static; new studies continuously update the evidence base. Ahuja’s framework is designed for what researchers call living data. Rather than rebuilding a table from scratch each time a fresh article appears, the system can incrementally adjust existing entries, preserving consistency while integrating novel findings. This capability is especially valuable for living systematic reviews or clinical decision support tools that must stay current without incurring prohibitive re‑processing costs.
Industry Recognition and Academic Mentorship
The rigor and practical relevance of Ahuja’s approach caught the attention of IBM, which awarded him the Infrastructure Master’s Fellowship Award—a honor reserved for research demonstrating strong real‑world impact. Vivek Gupta, assistant professor of computer science and engineering and head of ASU’s Complex Data Analysis and Reasoning Lab (CoRAL), praised Ahuja’s contribution: “Naman’s work really captures what we’re trying to do in CoRAL. We’re focused on complex structured data, especially how to generate it and evaluate it correctly, so we can build AI systems people can trust in real‑world settings… The IBM fellowship is well‑deserved.” Gupta’s mentorship helped Ahuja navigate experimental setbacks and refine both methodology and evaluation metrics.
From Hyderabad to the Seattle Tech Scene
Ahuja’s journey began in Hyderabad, India, where he earned his undergraduate degree in computer science before enrolling at ASU in 2024. During his graduate studies, he served as a teaching assistant for a natural language processing course, delivered a guest lecture on neural networks, and presented his research at an international conference in Vienna. Outside the lab, he maintains balance by playing basketball, exploring new music, and unwinding with stand‑up comedy in Hindi and English. Having accepted a full‑time role at Amazon in Seattle, Ahuja aims to work on “the core systems, how these models are actually built and how they can be used to solve real‑world problems,” aligning with his long‑standing goal of translating academic research into industry‑scale solutions.
The Growing Need for Usable Knowledge from Text
Even as Ahuja prepares to embark on his professional career, the fundamental challenge he addressed remains urgent. The world is producing more textual data than ever—research papers, clinical trial reports, health‑record notes, and online resources—yet much of it stays locked in unstructured form. Until AI can reliably turn that raw text into trustworthy tables, “someone, somewhere, [will still be] building a table by hand.” Ahuja’s incremental, traceable pipeline offers a promising pathway toward automating that labor‑intensive step, potentially freeing clinicians and analysts to devote more time to interpretation, insight, and ultimately, better patient outcomes.
https://news.asu.edu/20260430-sun-devil-community-computer-science-grad-earns-ibm-fellowship-ai-research