alex@nyc:~$

Alexander Xie

Software engineer at IBM. Into inference, interpretability, post-training, safety, optimization, multi-agent orchestration, and infra.

The bars are what I'm into lately: each run samples one. Edit the prompt and press Enter to try your own; drag to rotate.

$ cat experience.log

Experience

IBM
layer 05

Software Engineer

IBM · New York, NY
Jun 2026 – Present
  • Building AI agents and data pipelines on the Nexus Squad (Agentic AI & Workflow Automation)
Python · watsonx.data · Spark · Presto · dbt · Airflow · Confluent · Langfuse
Amazon
layer 04

Software Development Engineer Intern

Amazon · Boston, MA
Jun 2025 – Aug 2025
93% faster config creation
  • Built a Bedrock AI agent that automated config generation, reducing creation time by 93% for 400+ developers
  • Engineered a data pipeline using Lambda, DDB, and OpenSearch to auto validate and process 3K+ configs per day
  • Published agent to Amazon's package manager with evaluation layer, boosting config correctness by 40% org-wide
  • Scaled adoption across PXT as teams integrated package to streamline workflows and cut config turnaround by 75%
Bedrock · Lambda · DynamoDB · OpenSearch · CDK
layer 03

Machine Learning Research Intern

ColAI · Remote
Jul 2024 – Dec 2024
+18% forecast accuracy
  • Automated ETL pipelines for ingest, cleaning, and feature engineering in PyTorch, cutting runtime by 50%
  • Developed transformer forecasting pipeline across 30 datasets, improving accuracy 18% and anomaly detection 22%
PyTorch · Python · Pandas · Scikit-learn
New York University
layer 02

Student Computer Technician and Developer

New York University · New York, NY
Feb 2024 – May 2026
45% fewer misconfigurations
  • Automated security configuration enforcement in Bash across 700+ systems, reducing misconfigurations by 45%
  • Built monitoring workflows for network reliability, enabling early detection of failure and cutting downtime by 30%
  • Designed Tisch IT intranet into internal dev platform, improving tool discoverability and team efficiency
Bash · Linux · Python · Networking
Fortinet
layer 01

Software Application Intern

Fortinet · Sunnyvale, CA
Jun 2023 – Aug 2023
400+ vulnerabilities fixed
  • Detected and remediated 400+ vulnerabilities, preventing XSS, SQL injection, and privilege-escalation attacks
  • Reproduced 150+ suspected flaws in sandbox environment to verify exploitability and confirm patch success
  • Partnered with InfoSec to improve scanning coverage and coding patterns, increasing protected codebase by 20%
Python · Burp Suite · SAST · SQL
$ ls projects/

Projects

Real-time voice agentSept 2025 – Present

Orbit

A real-time AI technical interviewer you talk to. It listens, keeps track of the conversation, and answers out loud.

Architecture
LiveKit audio → speech-to-text → Gemini (Vertex AI) → ElevenLabs TTS → LiveKit audio
LLM calls run in a dedicated inference worker, separate from the Next.js app, so real-time audio never waits on the UI.
How it works
Context-aware prompting with conversation memory, so it gives structured, interview-style guidance instead of one-off answers.
Stack
Next.js · TypeScript · LiveKit · Vertex AI (Gemini) · Docker
Agentic code reviewFeb 2025 – Present

Aura

A GitHub App that records a video of your UI change actually working and posts it on the pull request.

Architecture
PR webhook → Gemini reads the diff → writes a Playwright script → sandboxed run → video posted to the PR
How it works
Reviewers get visual proof instead of “trust me, the button works”, with tests generated from the diff rather than written by hand.
Stack
GitHub Apps · Gemini 1.5 Flash · Playwright · Next.js

More projects

LLM Inference Engineering

An 8-week workshop on LLM inference: KV caching, FlashAttention, batching, quantization, and serving, with lecture notes and Jupyter labs.

PyTorch · CUDA · Jupyter
Apr 2026 – PresentGitHub ↗

Turtl.Bio

Regulatory intelligence for biotech: RAG over 1,000+ FDA documents, shaped by 15+ customer discovery interviews.

System Architecture · LLMs · Retrieval-Augmented Generation · API Design
Aug 2025 – PresentGitHub ↗

SignalFeed

ML news aggregator that scrapes arXiv, Hugging Face, RSS, and X every 6 hours, dedupes, and summarizes locally with Ollama.

Python · FastAPI · Next.js · PostgreSQL · Ollama · Docker
Mar 2026 – PresentGitHub ↗

Gauge

Training dashboard over Hevy and WHOOP data with ACWR, plateau detection, and local-LLM coaching.

Next.js 16 · TypeScript · Ollama · PostgreSQL · Tailwind CSS
Mar 2026 – PresentGitHub ↗

ASLingo

Real-time ASL alphabet recognition from 21 MediaPipe hand landmarks, classifying 26 letters with TensorFlow.

Python · TensorFlow · MediaPipe · OpenCV
Feb 2025 – Aug 2025GitHub ↗

Chip8 Emulator

All 35 opcodes in C++, a 64×32 framebuffer at 60 Hz via SDL2, with register and memory debug tools.

C++ · SDL2 · CMake · Emulation
Jan 2025 – Feb 2025GitHub ↗

Education

New York University
Dual Major in Computer Science and Data Science
Aug 2022 - May 2026

Coursework: LLMs, Software Engineering, Machine Learning, Algorithms, Data Management, Operating Systems, Causal Inference

Skills

Languages

C++C++
PythonPython
TypeScriptTypeScript
GoGo
JavaJava
CC
SQL
SQL
JavaScriptJavaScript

Frameworks & Tools

PyTorchPyTorch
TensorFlowTensorFlow
ReactReact
Node.jsNode.js
MED
MediaPipe
OpenCVOpenCV
PandasPandas
Scikit-learnScikit-learn
LinuxLinux
BashBash
SparkSpark
AirflowAirflow
DBT
dbt
LAN
Langfuse

Cloud & Infrastructure

AWSAWS
DockerDocker
PostgreSQLPostgreSQL
Google CloudGoogle Cloud
RedisRedis
MongoDBMongoDB
CLO
Cloudant
IBM
IBM Cloud
WAT
watsonx.data
PRE
Presto
APA
Apache Iceberg
Confluent KafkaConfluent Kafka

GitHub Activity

View Profile
MonWedFri