Skip to content

$ whoami

→ alex kang · governments call me "zixuan"

Hi, I'm Alex — a software engineer and ML researcher

CS grad from UBC, incoming MSCS at Columbia. I build systems that move money, models and infrastructure — low-latency Rust trading engines, transformer fine-tuning, data pipelines over 50M+ records, and agents that ship real work.

Scroll DownRésumé
Alex Kang
Currently buildingFizz.bot_
Fall seven times, stand up eight.
— Japanese proverb

#projects

clawlet — zsh

$ clawlet buy --item 'groceries' --usdc 60

→ wallet CLaW7f…q4CLaW · 214.30 USDC

→ policy ok · tx 60.00 ≤ cap 100.00

✓ settled in 1 block · card redeemed

TypeScript · Next.js · Solana · Bun · Prisma

Clawlet

Tell an AI agent what you need and it buys it for you. Real-world goods, paid in crypto, inside spending limits you set — no checkout, no card.

Code
stackbrief — zsh

$ stackbrief ingest --since 15m

→ 383 sources · 1,204 articles

→ embedding + clustering … 240 clusters

✓ 96 duplicates collapsed · published

Next.js 16 · TypeScript · Supabase / pgvector · Claude · Voyage AI

StackBrief

The fastest way to keep up with AI. One brief instead of 130 feeds — every story appears once, ranked by how many outlets covered it rather than by engagement, so it can't be gamed.

LiveCode
infra-designer — zsh

$ infra design && infra audit

→ spec.json → alb + ecs + rds + s3

→ terraform plan: 14 add, 0 change

! drift: rds.class m5.large ≠ t3.medium

Python · Terraform · AWS · MCP

infra-designer

Describe what you're building and get cloud infrastructure that fits. Your coding agent plans it, ships it, and tells you when reality drifts from the plan.

Code

#experience

Data Science Intern @ Siemens

May — Sep 2025

Beijing, China · Python · PySpark · Keras · AWS

  • Built a business classification system — fine-tuned transformer plus LightGBM, with focal loss for class imbalance — that improved F1 by 15% over the prior model.
  • Built a PySpark pipeline consolidating 50M+ customer pricing records into annual product- and customer-level price-change metrics.
  • Added a Keras VAE that flags anomalies in those pricing records, plus a Tesseract OCR pipeline that pulls contract numbers out of scanned PDFs by spatial alignment and binary search.

Co-Founder @ Ace Automations

Dec 2023 — Jan 2025

Remote · Rust · Tokio · gRPC · Solana

Users generated 7 figures in profits, including a single $3M day on $MOTHER

  • Led development of a profitable meme-coin trading system on Solana — the first automation software built on pump.fun.
  • Built a multi-threaded transaction handler on a thread pool that fills buy/sell orders in under 100ms, roughly 20× faster than manual execution.
  • Designed the gRPC and protobuf layer streaming to distributed clients, plus a statistical detector that flags anomalous launch patterns to keep users out of rug pulls.

Software Developer Intern @ LMI Technologies

Jan — Sep 2023

Burnaby, BC · C++ · C · Python

  • Rewrote sensor initialization as a multi-threaded pipeline, cutting start-up time by 40%.
  • Cut 20MB off build artifacts by resolving technical debt, and fixed the intermittent multi-threaded build failures that had been destabilizing the team's build server.
  • Built the C++ mocking infrastructure the team now tests against.

#research

Railway AI Video Surveillance Researcher @ Peking University

Jun — Nov 2025

Supervised by Prof. Zhang · Multimodal LLMs · LoRA

  • Fine-tuned multimodal LLMs with LoRA adapters and a temporal sliding-window sampling strategy to detect mudslides from live railway video and sensor feeds.
  • Built environmental augmentations — rain intensity, fog density, debris — to cut false positives and harden recall across weather conditions.

Researcher @ Algoverse AI Research

Jun — Nov 2025

Mentored by Shreyas (Amazon) & Kevin Zhu (UC Berkeley) · LLM reasoning · Bayesian methods

Accepted to the Agentic AI in the Wild workshop, ICLR 2026

  • Extended the Uncertainty of Thoughts framework for LLM information-seeking with Bayesian prior estimation.
  • Added multibranching to generalize beyond binary entropy-based splitting, plus likelihood caching over sentence-clustered questions.

#skills

C++c++
Rsrust
Pypython
TStypescript
Solsolana
SQLsql / nosql
Ptpytorch
Tftensorflow
Skscikit-learn
Sppyspark
Nxnext.js
Rereact
Awsaws
Tfmterraform
Dkdocker
RNreact native

$ cat skills.txt

languages
C++, Rust, Python, TypeScript/JavaScript, Java, SQL/NoSQL, x86-64, CUDA, Erlang, PHP
machine learning
Transformers, LoRA fine-tuning, VAEs, CNNs, SGD, MLE/MAP estimation, PCA/t-SNE
frameworks
PyTorch, TensorFlow, Scikit-learn, Pandas, NumPy, PySpark, OpenCV, React Native, Next.js
infrastructure
Kafka, Docker, Terraform, AWS (Lambda, ECS, VPC), PostgreSQL, Redis, gRPC/protobuf
ai tooling
Claude Code, Codex, Cursor, MCP servers, agent skills, prompt & context engineering
interests
Blockchain & DeFi, fintech, prediction markets, LLM reasoning & alignment, vision research

#about-me

$ cat about.md

Hey — I'm Alex. Zixuan Kang is my legal name, so that's what shows up on my email and paperwork. I just finished my CS degree at UBC and start my Master's at Columbia this September. Most of what I build lands in fintech, crypto and AI — systems where being fast and being correct are the same requirement.

So far that's meant co-founding a crypto trading company, data science at Siemens, systems work at LMI, and LLM reasoning research. The receipts are up in experience.

Right now I'm building Fizz.bot — all of crypto on your fingertips. Spot, perps, memecoins and prediction markets in one place, with a social layer that puts real skin in the game. The web app is live today; we're pivoting to mobile next.

I keep coming back because I genuinely love this — I trade crypto myself and can sit with it for hours without noticing. Years of that add up to real domain knowledge and a network of people in the space, and neither is something you can cram for.

When I'm not shipping, I'm in the gym or on a basketball or tennis court. Staying fit comes above all else.

$ cat education.txt

  • Columbia University

    M.S. Computer Science

    New York, NY · Starting Sep 2026

  • University of British Columbia

    B.Sc. Computer Science

    Vancouver, BC · Sep 2020 — May 2026

    GPA 3.76 / 4.0

relevant coursework

Advanced Algorithm Design · Advanced Machine Learning · Parallel Computation · Computer Hardware & Operating Systems · Advanced Relational Databases · Software Engineering

#contact-me

I'm open to new grad and internship roles, research collaborations, and anything at the intersection of ML and low-latency systems. The fastest way to reach me is email — I reply to everything.

Heads up: my email is under Zixuan Kang — my legal name. Same person as Alex.

Download résumé

$ contact --list