Kaushik Kallam

Selected work

Each case starts with its question and my part in it, then explains the research decisions and ends with what the evidence can and can’t show.

  1. Data Instrumentation Coverage and Quality

    When a product records an interaction, does that record actually explain what the person did?

    JPMorganChase · PaymentsNew York2026, ongoing

    Contribution
    I created the dashboard and the assessment pipeline.
    Methods
    • Instrumentation coverage audit
    • Review of tag meaning and context
    • Cross-product comparison with product drill-down
    Status
    Dashboard and pipeline created · being publicized

  2. Chegg Discord

    Could academic support fit into the way students already study together on Discord, and which AI-assisted capabilities were ready to develop further?

    Chegg · Discovery and concept evaluationRemote from TexasEight weeks, 2024

    Contribution
    I ran the discovery research and the concept tests.
    Methods
    • Two rounds of unmoderated discovery interviews with 24 STEM students
    • Three low-fidelity concept tests, 12 STEM students each
    Status
    Discovery and concept testing completed · recommendations for alpha and iteration

  3. Chegg Mexico

    What did localizing a learning product for students in Mexico require beyond translation?

    Chegg · Mixed-methods localization researchSanta Clara County, CaliforniaSix weeks, summer 2024

    Contribution
    I worked across the survey and interview phases and made the cross-language search recommendation.
    Methods
    • Comparative survey of 1,000 students, 500 in the US and 500 in Mexico
    • Unmoderated interviews with 12 English-speaking college students in Mexico
    Status
    Mixed-methods study completed · cross-language search implementation reported

  4. Inspire Brands Ad Creative

    Which creative attributes of quick-service restaurant TV ads go with higher ACE Metrix scores?

    Inspire Brands · Quantitative advertising researchAtlanta, GeorgiaSummer 2023, readout on August 8

    Contribution
    I defined the attributes with the head of Demand Gen Analytics, coded all 548 ads, ran every regression in R and presented the readout.
    Methods
    • Content coding of 548 quick-service restaurant TV ads from the past year on 21 yes-or-no attributes
    • A codebook defined before any ad was watched, with counting rules for ambiguous attributes
    • Linear regression of the Overall ACE Score and its seven components on the coded attributes, in R
    Status
    Analysis completed · readout presented August 2023, used to inform creative guidance

  5. watched.

    How can people express what they really think of a film or show, without ranking it feeling like work?

    Cofounder · Personal project2025 to 2026

    Contribution
    Cofounder: research, product and interaction design, and front-end work, with two technical cofounders.
    Status
    Launched project · currently paused

Also on the journey