Welcome.
I’m a builder. I’ve worked as a product manager and software engineer for over a decade.
hello@ankushagrawal.com2021–2026 · Product leadership
For members, I built products that made it easier to choose what to watch. Behind the curtain, I built systems that helped creative teams decide how each title should introduce itself.
Living room
01 · Conversational TV
I built Netflix’s first conversational TV experience. Ask about the character on screen, catch up on last season, or find something you and your spouse might both enjoy tonight. It had to feel less like search and more like Netflix understood the assignment.
The Verge
02 · Netflix homepage
I led the product experience for evaluating titles in Netflix’s new TV homepage: what to show, which evidence earns attention, and how quickly someone can decide whether to press play.
The New York Times03 · Decision evidence
I owned the systems for artwork, video, synopsis, badges, and callouts: the evidence members use to decide what deserves an evening. The work redesigned how Netflix explains a title before the first frame plays.
04 · Search & evaluation
I owned Search and the evaluation systems we used to tell whether discovery was getting better. One feature changed the artwork in a search result to reflect the query, so a person, relationship, mood, or theme could make the reason for the match immediately clear.
Behind the curtain
01 · Muse
I changed our focus from creating artwork at scale to creating better promotional assets through creative insights. Muse paired creative judgment with evidence from how members responded, giving artists and strategists a shared language for what worked and why.
Netflix TechBlog02 · Personalization & learning
Netflix can introduce the same title differently to different members. Our work connected that personalization to continuous learning: creative teams defined the possibilities, and member response helped reveal which artwork and video ideas connected with different audiences.
03 · Creative feedback
Muse showed which faces pulled people in, which relationships mattered, when action beat atmosphere, and how response changed across audiences. Underneath it, trillions of rows became insights people could use without writing a query.
How Muse works04 · Iris + Artwork Portal
I owned Iris and Artwork Portal, the operational backbone for requesting, scheduling, tracking, and delivering promotional artwork across titles, languages, and markets. I concentrated a long-running program, cut scope, and got it over the finish line.
05 · AVA
I owned AVA, the workspace where artists and agency partners could search a title’s visual source material, find and save promising stills, see similar assets, and open selected frames in downstream creative tools.
06 · Clips
I owned Clips, the workspace for finding and selecting useful moments from video. Starting from a CLIP-based base model, anyone could train a specialized model with as few as 100 labeled examples, then use it to surface the kinds of moments their workflow needed.
2015–2020 · Engineering → Product
Uber is regulated city by city, not from one national rulebook. I led the identity and onboarding systems that turned millions of local requirements into a path people could actually complete.
01 · Engineering leadership
I led the engineering work across document capture, classification, transcription, fraud detection, human review, secure storage, and configurable checks. These systems powered identity and onboarding across Uber’s global marketplace.
Public architecture from a later evolution
02 · Immediate feedback
I changed failure prevention from a slow review outcome into immediate feedback. Drivers could fix common document problems while the task was still in front of them instead of discovering them after a failed review.
See the later client-side implementation03 · Financial compliance
I led implementations that translated European KYC and AML requirements into onboarding checks, review paths, and operational controls. The work kept Uber compliant without turning every legitimate driver into an exception case.
04 · Regulated data
I owned the processing and storage systems for more than 10 million driver and rider documents per month, including Uber’s largest repository of regulated PII. Encryption, access controls, retention, deletion, and auditability were product requirements.
05 · Regulatory infrastructure
Most industries are regulated nationally or by state. Ridesharing can change city by city and job by job. I owned the instruction systems for millions of job-and-city combinations, so onboarding could tell each person exactly what applied where they wanted to work.
06 · Greenlight Hubs
I helped launch the first version in one week from a hotel room in 2015. It became Uber’s global scheduling layer for Greenlight Hubs and generated $10M+ in annualized incremental revenue.
Read about the Greenlight Hub systemThe Sudoku app I always wanted: ad-free and offline-friendly, with 500 certified puzzles that never require guessing, ratings based on verified logical solve paths, progressive coaching, and lessons plus three practice modes for 17 techniques. I’m also working on a research paper that uses computational experiments to measure how much additional solving coverage advanced techniques provide.
A platform that connects what’s changing in the world to people trying to make better decisions—discovering and monitoring entities, ranking what matters, supporting action, tracking outcomes, and learning without surrendering domain authority.
Personal agents can unlock powerful workflows, but installing and configuring Pi, Hermes, or OpenClaw is still intimidating for many people. I’m building an agentic app compiler that lets developers turn those workflows into packaged Mac apps, choose local models, an existing AI subscription, or a paid API, and ultimately offer users a signed app with one-click download and installation.
A public-data-first tool that helps San Francisco homeowners find comparable sales and decide whether a property-tax assessment appeal is worth filing.
A mobile narrative-album builder that turns a large photo set into editable story sections, captions, and shareable versions for different audiences.
I’m building Bumi Commons, a small nonprofit for overlooked corners of the city I call home. Its first project is Lyell Hill, where neighbors have reclaimed sidewalk, cleared overgrowth, worked with Public Works, and begun a community mural between Glen Park and Bernal Heights.
An installable Python package that preserves Tuya and Smart Life temperature and humidity history beyond the cloud’s seven-day window. It deduplicates and corrects readings, pseudonymizes raw device IDs, archives to SQLite, JSONL, or CSV, exports hourly summaries, and can schedule daily collection on macOS.
A local control surface for recurring personal automations, with one-click reruns of curated workflows, live scheduler state, an exception-focused attention queue, and safeguards against duplicate runs.
A library skill that maintains a ranked reading queue, checks SFPL copy-level availability, estimates transfer and return timing, and alerts you when to place the next hold so one book is ready as another ends.
A private pool for four friends, with live standings, transparent scoring, and a full comparison of everyone’s picks. Preserved here as a final, read-only snapshot.
I spent years as a Room Advisor, working closely with high school students aiming to become the first in their families to graduate from college. It remains one of the clearest examples in my life of how much it matters to keep showing up.
I consulted for enterprise SaaS companies building Kafka products. As a developer and expert user, I helped their teams improve the experience for other developers and communicate the product’s value to less technical audiences.
I studied computer science at UC San Diego, fell hard for La Jolla, and now live in San Francisco.
I spend a lot of time outside. I hike Mount Shasta every year, and in 2025 I summited via the Hotlum-Bolam Ridge route. I return to the Lost Coast Trail every year, too. I like backpacking and multi-pitch rock climbing because they keep teaching me to prepare carefully, stay curious, and carry real responsibility for the people around me.
I love a great story. I love telling one, and I love hearing yours. That instinct follows me into product work, teaching, mentoring, and community work. The best stories reveal what people care about, where a system is failing them, and what might be worth building next.