PRODUCT × ENGINEERING × AI

SIJIA·HU

build · think · repeat

A product thinker and engineer in the age of AI

I break down complex problems and build solutions that work

Portrait of Sijia
SCROLL ↓

About

I care about whether a problem is worth solving

I studied computer science at the University of Wisconsin–Madison, but three real problems taught me the most: at DataCEVA, I brought workflows scattered across Slack, Jira, and email into one platform; at OfferContext, I built a membership growth and settlement flow that could be reconciled end to end; and for my own AI email assistant, I interviewed 42 international students, then cut half the features I had planned.

Each time, I followed the same approach: define the problem and the measure of success before building. Weekly report preparation dropped from 90 to 40 minutes, and email classification accuracy rose from 89% to 95%. Those were measures I chose up front to test whether the work was useful.

Decide how to prove it helps before writing the first line of code

Capabilities

What I can do

Four kinds of work I can deliver, from problem framing to a working product.

01

Frame the problem

Find the real need through user interviews, turn it into a practical workflow and priorities, and define success measures before building.

User interviewsRequirementsWorkflow designMetrics
02

Ship AI features

Bring language models into real workflows through structured skills and prompts, RAG retrieval, and conversational memory. Use evaluation sets to check each change.

LangChain4jRAGPrompt / SkillEvaluation
03

Build the backend

Take an idea to a deployable service, including APIs, access control, data consistency, and correctness under concurrent load.

JavaPythonSpring BootFastAPIMySQLRedis
04

Move a team forward

Break down goals, assign work, and make decisions with limited time and incomplete information. I work comfortably in English across technical writing and conversation.

PlanningCross-functional workTOEFL 99

Projects

Things I've built pinned up

The problem, who it affected, and the result.

OfferContext Membership Growth & Monetization cover image
01 / 03

OfferContext Membership Growth & Monetization

Product & Backend Development · Beijing Qiuye Technology

Built referrals and paid conversion for a graduate job-search product. The system tracks who invited whom and reconciles rewards, payments, and benefits. Limited-benefit claims had zero overselling and zero duplicate claims under high concurrency; the beta claim conversion rate was 12.6%.

JavaSpring BootRabbitMQRedis
AI Email Assistant cover image
02 / 03

AI Email Assistant

Product Lead · Independent Project

Helped international students pick out the few messages that matter from more than 20 emails a day. A digest arrives in the chat app they already use, with no new app to install. In a two-week pilot with 22 users, classification accuracy rose from 89% to 95%, daily handling time fell from 25 to 18 minutes, and two-week retention reached 82%.

User interviewsLLMFeedback loop
OfferContext AI Workbench cover image
03 / 03

OfferContext AI Workbench

Skill Design & RAG Q&A · Beijing Qiuye Technology

Built an iterative workbench for resume reviews and job-description matching. Structured skills make the criteria explicit; a RAG knowledge base supplies evidence. With an evaluation set in place before prompt changes, resume/JD matching F1 rose from 62% to 78% while tokens per review fell by about 30%.

LangChain4jRAGSkill / PromptEvaluation

Experience

The path so far

Mar 2026 — Present

Product & Backend Developer

Beijing Qiuye Technology · OfferContext

Owned requirements review, business rules, and backend implementation for membership growth and monetization, from referrals through reward settlement. Also worked on AI workbench skills and RAG Q&A. Evaluation-driven changes raised resume/JD matching F1 from 62% to 78% and reduced tokens per review by about 30%.

Aug 2025 — Dec 2025

Software Development Intern

DataCEVA · Chicago

Helped build an internal collaboration and ticket-management platform from scratch, working across frontend and backend. Owned collaboration workflows and request management, and helped introduce AI-assisted ticket creation. The first release served more than 20 internal staff and their clients; weekly report preparation fell from about 90 to 40 minutes.

Sep 2022 — Jan 2026

B.S. in Computer Science

University of Wisconsin–Madison · School of Computer, Data & Information Sciences

GPA 3.6/4.0 (top 15%); TOEFL 99. Led a seven-person team through timed challenges at the Nanjing round of the Newth National Youth Challenge, setting goals, assigning roles, and making decisions on site. We placed first overall among 12 teams and 84 participants.

Thoughts

A few things I believe

The ideas I expect to discuss in an interview, laid out here first.

01

AI won't replace everyone who writes code

It will take over work that is purely execution. Judgment, trade-offs, and defining the problem become more valuable.

02

Make it work, then make it elegant

A rough product people can use is worth more than a beautiful proposal. Real feedback comes from real use.

03

Clear writing is a technical skill

If I can't explain something clearly, I probably haven't understood it well enough. Documentation, commits, and conversations draw on the same skill.

Let's talk about anything

Hello there,nice to meet you.

I warm up slowly, but I can talk for hours about a good idea.

Talk shop

New ideas, projects,or possibilities

Grab coffee

Online or in person,either sounds good

Swap playlists

Here's one songyou have to hear

Send me a note ↗