Rachael Chew

Work & Products

Things I've shipped

Each product is built for a specific audience, solving a real problem at the intersection of AI and human understanding.

Comms Teams

Pulse

I wanted to understand how narratives about AI form and spread — so I built the tool I would want if I'm on the communications teams had. Pulse monitors media sentiment, creates tailored briefs, tracks narrative drift across Discord and Twitter, and fires Slack alerts when something breaks. I couldn't find a tool that synthesises news about anything I am interested in, so I began to build this.

Next.jsSupabaseClaude APISlack Webhooks

Watch Demo

Real-time narrative intelligence across platforms

AI Safety Researchers

SYSCONBENCH

Everyone says AI models shouldn't be sycophantic. I wanted to measure how much they still are. SYSCONBENCH tests frontier models across six behavioral categories and found that even post-RLHF, models mirror user values 25.9% of the time. I grounded each part of the evaluation framework on five branches of cognitive science research, from Dunning-Kruger calibration to Kunda's motivated reasoning. I wanted to know not just how much models agree, but why they can't stop.

PythonTransformer LensEval Framework

GitHub

25.9% residual value-mirroring drift discovered

Explorers & Travelers

Buddy

A place-saving and guide-sharing app I built solo from zero. AI-generated city guides, semantic search, a taste graph that learns what you love, and trip journaling with time-travel. The kind of app I wanted to exist, so I made it.

React NativeExpoSupabaseClaude SonnetMapbox

GitHub

27 days from first commit to full product, solo

Cancer Researchers & Clinicians

GenomIT

At genomIT's CAETI Research Group in Buenos Aires, I built modular Python pipelines processing 10,000+ genomic samples and applied unsupervised ML — K-Means, Spectral Clustering, OneSVM — to 30,000+ breast cancer biomarkers. Ran Kaplan-Meier survival analyses and Cox proportional hazards modeling. Shipped a Dockerized dashboard used by 40+ clinicians. For the first time, my work met the illness that took my mum.

Pythonscikit-learnK-MeansKaplan-MeierDocker

GitHub

30,000+ biomarkers analyzed, patterns used in 21% of prognosis models

Visually Impaired Users

EchoAid

Two medication bottles that look identical except for their labels. One wrong choice could be dangerous. At Black Wings Hackathon 2024, we built EchoAid — an NFC app that lets visually impaired users attach voice-tagged stickers to everyday objects and identify them with a tap. We didn't start with the technology. We started with user stories. Winner, Most Innovative Hack — 1 of 78 teams.

NFCReact NativeVoice UIAccessibility

Watch Demo

1st Place Most Innovative Hack, Black Wings 2024 — 78 teams

Youth in Singapore

Mental Health Collective

At 18, I co-founded Singapore's first nationwide youth mental health initiative. No template, no institutional backing. We built government partnerships from cold outreach, designed a conference system that scaled to 3,000+ attendees, and created infrastructure that outlasted us. This is the project that taught me what building for people actually means.

Community DesignPolicy PartnershipsEvent Systems

3,000+ lives reached, institutional infrastructure that endures

Where I've Worked

2025 — Now

Independent Builder

Pulse · Buddy · SYSCONBENCH

San Francisco

2024

DevRel Associate

Quantstamp / Zircuit

San Francisco

2023

Software Engineer

genomIT

Buenos Aires

2023

Software Engineer

Columbia Irving Institute

New York

2022

Engineer & PM

Hikingbook

Taipei

2018–2022

Co-Founder

Mental Health Collective

Singapore

The Journey

8 cities. 4 continents.
Synthesizing Chaos.

Every four months, a new city. New language, new systems, no playbook. Two suitcases and one question: what can I build here? After eight times, I stopped fearing blank pages.

1.3521° N, 103.8198° E

Singapore

Home base · Where it all started

Co-founded Mental Health Collective with 4 friends. Ran COVID operations at the Ministry of Health coordinating 40,000+ patient workflows. Learned that systems don't scale unless you design for the humans inside them.

Home

Building

37.7749° N, 122.4194° W

San Francisco

Current · Building

Where code meets conviction. Building Pulse, Buddy, and SYSCONBENCH. Learning to ship fast and think slow about safety.

37.5665° N, 126.9780° E

Seoul

Lived with a korean family!

Studied the tension between rapid tech adoption and cultural preservation. Learned that speed without wisdom is just noise.

Reflection

Narrative

51.5074° N, 0.1278° W

London

DevRel @ Quantstamp

Ran developer relations for Quantstamp's Build To Earn program. Built a submission triage pipeline for 165+ entries, onboarded 200 partner teams across four program rounds. Learned community-building is a product in itself.

52.5200° N, 13.4050° E

Berlin

Minerva rotation

A city that rebuilt itself from scratch. Made me think differently about what it means to build something that lasts versus something that just works.

Rebuilding

Scale

17.3850° N, 78.4867° E

Hyderabad

Interned at a cognitive science lab

Studied how cultural context shapes experimental methodology. Realized that the same research question gets unpacked completely differently depending on the lens we choose to take on.

25.0330° N, 121.5654° E

Taipei

Hikingbook · Software Engineer & Product Manager

Wrote production code in Mandarin. Product-managed feature decisions using SQL analytics on 20,000+ user events at Hikingbook. First time I understood that data doesn't speak — you have to interrogate it in the right language.

Curiosity

Discovery

34.6037° S, 58.3816° W

Buenos Aires

Cancer Research @ genomIT

Built genomic data pipelines at genomIT. Applied unsupervised ML to 30,000+ cancer biomarkers, surfacing patterns used in 21% of downstream prognosis models. Watched Messi live! Ate empanadas and alfajores at midnight. One of the best four months of my life.

About

Why I Build

Philosophy taught me to ask why — to sit with a question until the question itself changes shape.

Cognitive science taught me how people actually think — not to assume how our minds work, but to understand the messy, biased, beautiful reality of human cognition.

Computer science gave me the tools to build — to take those insights and ship something real by tomorrow morning.

That's why I build with doctors, researchers, teachers and more. Not just to study the intersection, but to make it work.

Cities where I've lived, worked, and shipped. I made myself a guinea pig of my own experiments on what life can look like. Each one changed how I think about building for humans across cultures.

54.5k

Lines of code in Buddy, a solo build sprint. I love to map ideas of tools and build them.

3,000+

Individuals reached through Mental Health Collective, the initiative I founded with 4 friends. My first product wasn't software, but community infrastructure.

Majors I studied at Minerva University — Computer Science and Cognitive Science. I chose both because the interesting problems live at this intersection.

The Interview Project

150 Changemakers. 45 Countries.

I wanted to understand what it means to live a meaningful life, so I went directly to the source.

CEOs, professors, authors, a butcher who became a vegan food founder. No warm intros. No platform. No budget. Just me, LinkedIn, and a question I couldn't let go of.

150

Conversations

Countries

42%

Response Rate

28%

Meeting Booked

The System

I built my first outbound playbook in 2020 before AI existed to help me. Cold outreach meant manual research, A/B testing subject lines in spreadsheets, and tracking response rates by hand. I tested 8 different message templates across 200+ prospects, iterating on length, personalization depth, and call-to-action framing.

The winning formula: 42% response rate, 28% meeting booked rate. Curiosity-driven outreach outperformed open-ended asks by 3x. Every conversation taught me something about building trust with strangers at scale.

What I Learned

The meta-lesson was about growth itself. How do you get busy, successful people to care about an unknown college student's questions? The answer wasn't luck. It was seeing every rejection as a data point.

20 people didn't respond to the 7-sentence version? Try 3 sentences. Still nothing? Change the subject line. Then send 100 more.

The people who'd figured something out weren't chasing credentials. They were obsessed with work that mattered to them.

By conversation 150, I had a conversion playbook that worked across cultures, industries, and seniority levels. The best training for growth work was hundreds of rejections and learning from each one.

Two suitcases, a laptop, and one question: what can I build here?