Anirudh KumarRésumé

Analytics leadership & applied AI

Clear thinking.
Useful things.

I turn messy questions into decisions, experiments and tools people can use.

Seven years in analytics at Intuit, LTV and Self Financial. Now bringing that same judgment to AI workflows and independently built products.

Download résumé Let’s talk

San Francisco Bay Area Open to the right problem

Built on experience at

Self FinancialLTVIntuitUC Berkeley

Selected experience

The work behind
the point of view.

I like the part where the question is still fuzzy.
Here’s what I’ve done with it.

Self Financial

Lead Growth Analyst

Apr 2025 – May 2026

From a conversion question to a testable decision.

I paired experimentation with an executive dashboard used near-daily to understand conversion and marketing shifts.

Read the work behind it
  • Built an executive SWAT dashboard used near-daily by CPO, VPs, and directors to diagnose conversion and marketing shifts.
  • Ran 10+ pre-paywall A/B tests across activation and conversion, shaping launch metrics with Product.
  • Built internal AI/analytics tooling: Slack and Confluence scrapers plus a Redshift data-dictionary generator.
  • Introduced LLM workflows into recurring analytics deliverables and helped teammates apply AI tooling.
10+

pre-paywall A/B tests

The Lifetime Value Company

BI Manager → Senior BI Manager

Mar 2021 – Jul 2024

Make good analysis a tool the team can use.

Self-service reporting, cohort exports and a consistent experiment framework helped teams answer recurring questions.

Read the work behind it
  • Built configurable self-service dashboards and cohort exports across dozens of metrics and filters for Product, Email, and Customer Success.
  • Built Python scripts to normalize and score messy vendor data proposals, including consensus checks when no source of truth existed; resolved schemas directly with vendors and advised the COO on purchases.
  • Increased Trustpilot review volume and ratings by building review scrapers and engagement segments for targeted outreach with Email and Product. The results reflected outreach to selected customers.
  • Owned the A/B-test evaluation framework that standardized experiment ramp and readout across BI.
  • Designed SQL interview assessments, conducted introductory interviews, reviewed resumes, and mentored three new hires.
$20M+

annualized brand revenue under analytics ownership

Intuit

Data Analyst → Senior Data Analyst

Feb 2019 – Mar 2021 · Intern, Summer 2018

Bring clarity to consequential operating decisions.

My work connected TurboTax service quality, headcount planning and the reporting systems behind Finance.

Read the work behind it
  • Owned TurboTax Full Service analytics for expert quality, scoring, error rates, experiment readouts, and Tableau reporting.
  • Built a $75M-favorability headcount forecast against the bottom-up plan and automated recurring finance/reporting workflows.
  • Administered Corporate Finance's Tableau site for 100+ users, including access controls, compliance, training, and office hours.
  • Analyzed tax-filer upgrade drivers to test whether product lineup and onboarding fit returning-user behavior.
  • Built risk-routing and quality analyses to surface high-rework return patterns across hundreds of tax fields.
$75M

forecast favorability versus the bottom-up plan

UC Berkeley B.A. Economics · GPA 3.95 · Aug 2015–Dec 2018Résumé DOCX

Independent products

Small friction.
Shipped fixes.

I define the problem, architecture and safeguards, then build with AI and iterate on the result. These are the tools that came out of that habit.

Live tool

PasteDrop

Share text with a link you can say

2-word URLs · 24h expiry Explore project →
Live tool

TrimText

Less filler. Clearer AI output

Local rules · visible edits Explore project →
Live tool

TypingReps

Practice the text you actually type

live WPM · weak-key analysis Explore project →
Explore all six projects
PasteDrop

Paste text, get a memorable two-word URL, share it, let it expire

Stack + role
Independently built with AI assistance. Next.js app with Neon, Redis rate limiting, expiry cleanup, and slug scoring

pastedrop.ai
Who it's for
People who need a quick, low-friction way to share disposable text
What problem
Most paste tools are either account-heavy or forgettable. PasteDrop makes the link human-readable and temporary
How it works
Create a paste, receive a two-word slug, share the URL, and let the cleanup job expire it
What I built
Core app, slug generation, password option, rate limits, expiry flow, and abuse-aware guardrails
TrimText

A structural compressor for verbose AI text: less filler, clearer structure

Stack + role
Independently built with AI assistance. Next.js app. Rules-based compressor first; optional LLM pass in the product roadmap

trimtext.dev
Who it's for
People who pass AI output into more AI, docs, emails, or notes and do not want to pay for filler twice
What problem
LLM prose tends to add scaffolding, hedges, recap framing, and softeners. That creates reading cost and token cost
How it works
A deterministic pass removes known bloat patterns while preserving markdown, URLs, numbers, commands, and facts
What I built
Compression rules, guardrails, UI, pricing/privacy pages, and a Chrome extension build path
TypingReps

Typing practice built around realistic work text instead of random drills

Stack + role
Independently built with AI assistance. React/Vite app with local history, WPM timing, and weak-pattern feedback

typingreps.com
Who it's for
People who write for work and want typing practice that feels like email, code notes, and technical prose
What problem
Most typing trainers optimize for arcade loops. TypingReps optimizes for one useful rep at a time
How it works
Pick a work-like prompt, type it through, then review WPM, accuracy, and weak patterns
What I built
The practice flow, timing model, prompt sets, local history, and feedback surface
TackNote

Captured in Chrome: create a sticky note, save it, then copy its text.

Stack + role
Chrome MV3 extension. Local/sync storage, quick copy, color labels, and prompt-site insert support

Chrome Web Store
Who it's for
People who reuse prompts, snippets, checklists, and small notes across AI tools and browsers
What problem
Useful text fragments disappear into notes apps, chats, and random tabs. TackNote keeps them one click away
How it works
Save notes in Chrome storage, pin or color-code them, search, copy, and insert into supported AI sites
What I built
MV3 extension, popup UI, sync/local fallback, quick insert command, and Web Store packaging
Refresh All Tabs

One toolbar click reloads Google and Yahoo Finance. Then explore skip rules, refresh scope, and advanced options.

Stack + role
Chrome MV3 extension. Tabs/windows APIs, context menus, options page, and local settings

Chrome Web Store
Who it's for
Power users who keep many tabs open and need a controlled refresh instead of twenty manual reloads
What problem
Reloading tabs one-by-one is repetitive, but blanket refresh tools often miss skip rules and cache behavior
How it works
Choose scope and skip rules, then reload tabs via popup, keyboard shortcut, startup option, or context menu
What I built
MV3 service worker, popup, advanced settings page, URL filters, delay handling, and Web Store package
Site Bulletin

Captured in Chrome: save a reminder for this site, then see it appear on the next visit.

Stack + role
Chrome MV3 extension. Optional host permissions, content injection, storage rules, and onboarding

Chrome Web Store
Who it's for
People who want a visible nudge on specific sites without hard-blocking themselves
What problem
Blockers are too blunt when the goal is remembering intent, not banning a site
How it works
Add a site/page rule, grant permission for that rule, then receive an auto-fading or sticky reminder when it matches
What I built
Rule manager, optional permission flow, Shadow DOM bulletin renderer, snooze behavior, and onboarding

The person behind the work

A systems thinker.
A hands-on builder.
Pogo’s older brother.

“At some point, shipping the fix becomes easier than continuing to tolerate the friction.”

That’s the thread from analytics to my own products. Work backwards from the problem, make the rules explicit, then build something that improves with use.

Outside work, I’m a Maltipoo’s full-time bodyguard, a math and reading regular, and drawn to anything that punishes sloppy thinking.

A little more about me

How I think

Most people see the world and life as a sequence of mysteries. I see it as a collection of patterns and systems. I've spent my career in analytics, my free time in finance and gaming, and most of my heart on animals. All three share one thing: working backwards from a problem, building rules that make improvement repeatable, then executing with discipline.

What I build

The side projects, including PasteDrop, TrimText, TypingReps, and three Chrome extensions, come from the same place: noticing a small friction enough times that the only way to stop seeing it is to ship the fix.

Away from the dashboard

Outside of all that, I'm a 14-year-old Maltipoo's full-time bodyguard and older brother, a math and reading regular, and drawn to anything that punishes sloppy thinking: strategic systems, fast-paced action, collaborative mystery-solving. On AI itself, I'm an optimist who actually means it. The current models are early, often clumsy, sometimes wrong. The trajectory is still the biggest shift happening right now, and I'd rather build inside it than narrate from the side.

What comes next

I'm looking for applied-AI and analytics leadership roles — where analytics depth meets hands-on AI systems. Remote preferred, San Francisco Bay Area hybrid open.

What I work with
Analytics
Python, SQL, A/B testing, product analytics, Google Analytics, behavioral and event data
Business intelligence
Tableau, KPI systems, self-service dashboards
Applied AI
LLM workflows, prompt engineering, RAG, multi-agent orchestration, MCP integrations
Cloud
AWS, GCP

Have a good problem?

Let’s make
something useful.

hire@anirudhkumar.ai

Remote preferred; open to Bay Area hybrid.
LinkedIn is the fastest way to reach me—I read every message and reply within a day.

Résumé PDFDOCXLinkedInGitHub
Ask about my work
Enter sends · Shift+Enter newline 0/1200
AI responses can be imperfect. How this chat handles your data

AI chat — responses are generated by configured models through Vercel AI Gateway, Groq, or Cerebras, with Llama 3.3 70B as the default, and may be imperfect. Chats may be logged with your message, a short response preview, time and latency, selected starter, provider route, broad device class, and a pseudonymous session ID. Raw IP addresses are not stored; do not enter personal or confidential information.