MY SEE, WHAT I SEE
Things I noticed and couldn't stop thinking about.
Sometimes it's data. Sometimes it's business. Sometimes it's technology.
Sometimes it's just me asking: "Wait... why does this work like that?"
Pick something I probably shouldn't have built
I initially assumed users abandoned their carts because prices were too high or coupon codes failed.
The City Is a Machine — Processing 37 Million NYC Taxi Trips with Medallion Architecture
I wanted to understand where an urban transportation network makes money, and more importantly: at what data scale does distributed PySpark actually become necessary compared to single-node DuckDB?
CASE STUDY: 110M Clickstream Events — What Actually Happens Between View, Cart, Remove, and Purchase?
I started with 109.9M e-commerce events and a simple question: can raw clickstream data tell us where customer intent breaks — and can we turn those observations into behavioral personas, Markov journeys, and a CEO-level what-if simulator?
CASE STUDY: Exploring E-Commerce Customer Behavior & Churn (50,000 Records) — Friction Signals, Spending Paradoxes, and Retention Realities
A personal exploration of a 50,000-customer e-commerce dataset, analyzing support friction signals, cart abandonment churn thresholds, spending paradoxes, and behavioral customer segmentation.
CASE STUDY: Exploring Global E-Commerce Sales (2021–2024) — Revenue Skew, Category Dynamics, and Data Limitations
A personal analytical exploration of 10,000 global e-commerce orders ($5.28M revenue), looking at revenue distribution, category and regional patterns, fulfillment behavior, and what the dataset can — and cannot — tell me.
CASE STUDY: Stack Overflow Developer Survey 2025 — Market Research & The AI Accuracy Trust Gap
An in-depth market research case study analyzing 49,191 developer responses across 177 countries from the Stack Overflow 2025 survey, exploring the gap between AI adoption velocity and developer trust.
CASE STUDY: Customer Segmentation & Insight Matrix — Transforming 15k Shopping Mall Records into Revenue Strategy
A consulting-grade data analytics case study: applying K-Means clustering (k=8) to 15,079 shopping mall customer records and bridging machine learning outputs with actionable business insight matrices (Finding/Evidence/Implication).
CASE STUDY: Why Doesn't Increasing the Database Connection Pool Always Increase Throughput?
A Senior-level interview case study about a deceptively simple production problem: your API is slow, so you increase the database connection pool — and the system gets even worse.
CASE STUDY: How Would You Design a Rate Limiter for a High-Traffic API?
A system design case study answering a deceptively simple backend interview question: how would you design a distributed rate limiter that remains accurate, fast, and scalable?
CASE STUDY: Kafka Guarantees Delivery — So Why Did My Consumer Process the Same Message Twice?
A Senior-level distributed systems interview case study about Kafka delivery semantics, consumer crashes, offsets, idempotency, retries, and why 'exactly once' is much harder than it sounds.
CASE STUDY: Your API Is Fast at p50 but Terrible at p99 — Where Is the Problem?
A Senior-level performance debugging case study: the average latency looks excellent, but a small percentage of requests take seconds. The tricky part is discovering why optimizing p50 may do almost nothing for p99.
CASE STUDY: Why Did Our API Get Slower After We Added Caching?
A senior-level backend interview case study about a counterintuitive production problem: adding Redis caching makes an API slower instead of faster.
CASE STUDY: Why Does Your AI Agent Keep Calling the Same Tool Forever?
A production AI agent can have a powerful LLM, excellent tools, and a solid RAG system—and still get stuck in an infinite tool-calling loop. This case study explores why that happens and how to design an agent that can actually control its own execution.