KarthikKumar

I lead AI product and strategy in life sciences. I also build and ship software myself, at work and on my own time.

  1. 01Clinical Trial Delta
  2. 02AI Medical Insights Platform
  3. 03Clinical Site Insight Accelerator
15-minute conversation See the work

01

Selected work

01of 09

Clinical Trial Delta

A daily tracker for clinical trials in GLP‑1/obesity, type 2 diabetes, and cholesterol. It shows what changed in each trial’s public record: status, enrollment, endpoints, and sites.

  • Public repo
  • Daily data pipeline
  • AI output tested
Clinical Trial Delta page for GLP-1 and obesity trials: 1,091 trials, 239 recruiting now, with charts of trials by phase and by status.
Fig. 01The GLP-1 and obesity page, listing every tracked trial in that group. Screenshot from a local copy of the app, with registry data from Nov 9, 2026.

02of 09

AI Medical Insights Platform

Took an AI insights platform from proof of concept to 477+ active users and 5x engagement, giving medical teams hours back every month.

  • Generative AI
  • RAG
  • Medical Affairs

Impact

Active users
477+
Questions per month
5x
Monthly active users
+75%
Hours saved per user per month
8.4

03of 09

Clinical Site Insight Accelerator

A clinical site insights platform I own at work. It brings in clinical site information from Veeva and shows study teams where problems are starting, at the global, study, country, or site level.

  • In production
  • Employer work

04of 09

AI Agents for Faster Drug Advancement Decisions

Designed a two-agent AI proof of concept to cut review cycles on post-Phase 1 go/no-go decisions.

  • Agentic AI
  • Drug development
  • POC

Success criteria

Target review cycles per case, from a baseline of about 2.3
<1.6
Target accuracy predicting reviewer issues
70%+
AI agents
2
Proof of concept
12 wk

05of 09

Financial Planning

A household financial planning app. It covers income, the budget, taxes, card benefits, and net worth, and a figure entered once is used everywhere it matters.

  • Live demo
  • Public repo
  • Fictional demo data
Financial Planning dashboard, marked Demo data: net income per month, savings rate, fixed costs, a bar showing where every dollar goes, a net worth chart, and a red panel flagging the most urgent item.
Fig. 05The dashboard of the public demo, which runs on fictional data. Screenshot from a local copy, Nov 10, 2026.

06of 09

Financial Research Hub

A stock research site built around a personal watchlist. For each stock it shows the price history, a score for how good a buying opportunity it looks, a valuation that helps judge the right time to buy, upcoming events, and news.

  • Live demo
  • Private repo
Research hub dashboard in demo mode: a watchlist of 25 tickers with price trend, percent off high, RSI, setup score and sentiment, beside upcoming catalysts and market headlines.
Fig. 06The watchlist dashboard of the public demo. Market data is a snapshot from Nov 9, 2026. Screenshot from a local copy.
  1. 07of 09

    Trip Hub

    A trip planner for a couple, a family, or a small group. It keeps a trip’s day-by-day plan, bookings, documents, budget, and packing lists in one place.

    • Live
    • Private repo
  2. 08of 09

    Stocked

    A shared grocery and household supplies app for a two-person household. It keeps store lists, pantry stock, and filter replacement dates in one place that both people see.

    • Live
    • Public repo
  3. 09of 09

    Recoup

    A tracker for the credits that come with premium credit cards. It shows which credits you have used, which are still open, which you missed, and whether the card is beating its annual fee.

    • Live demo
    • Private repo

02

How I build

Each app I build is where its data lives. AI agents do the work inside it, and I make the decisions that carry consequences.

  1. 01

    Written spec

    I write a build brief that says what to build and what not to.

  2. 02

    Agent build

    An AI coding agent builds against the brief.

  3. 03

    Acceptance criteria

    The brief lists the checks the build must pass. If it fails one, it isn’t done.

  4. 04

    My review

    I review the result before it ships.

Guardrails

  • Anything public runs on made-up data.
  • An agent needs my approval before any action with real consequences.

Artifacts

03

Career

  1. 2026–

    Senior Manager, Digital Product ManagementAmgen

  2. 2024–2026

    Manager, Digital Product ManagementAmgen

    Led an AI medical insights platform from proof of concept to scale across 17 product teams: 477+ active users, 5x questions per month, 8.4 hours saved per user per month.

  3. 2023–2024

    Manager, Digital Transformation & InnovationNovo Nordisk

    Ran 5 AI/ML university-partnership projects across rare disease and obesity. Rebuilt the innovation process to run 2x faster.

  4. Co-founderYTLS Health

    Wireless NICU monitoring. NSF I-Corps.

  5. 2017–2020

    Marketing ManagerYou First Services

    Took a biotech disinfection product from idea to launch, with pilots at Ohio State, Tennessee, and Buffalo, then full installation. Grew the product portfolio from $240K to $950K+ in a year.

  6. 2014–2017

    Clinical Research AssistantOishei Children’s Hospital

    Co-authored a Journal of Perinatology study.

MBA and MS Marketing, UT Dallas (2023). BS Biological Sciences, Buffalo (2015).