Skip to content
View coderloganli's full-sized avatar
🎯
Focusing
🎯
Focusing
4 followers · 3 following

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx2Achievement: Quickdraw

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx2Achievement: Quickdraw

Block or report coderloganli

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
coderloganli/README.md
Full-Stack AI Engineer — LLM / Agent Applications Open to work Location 
   ██╗      ██╗          logan@github
   ██║      ██║          ─────────────────────────────────────────────────────
   ██║      ██║          Status    : open to New Grad SWE / Applied AI  █
   ██║      ██║
   ███████╗ ███████╗
   ╚══════╝ ╚══════╝     Languages : Python · TypeScript · Rust · Java · Go · Swift · SQL
                         AI / LLM  : RAG · LangChain · LLM Evaluation · Function Calling
                         Frameworks: FastAPI · React · Spring · PostgreSQL · Redis · LiveKit
                         Cloud     : AWS Bedrock · Lambda · S3 · DynamoDB · Step Functions
                         DevOps    : Docker · GitHub Actions · OpenTelemetry · pytest · k6

💼 Experience

Career timeline — Jan–May 2026: Northeastern CESAR Lab, Machine Learning Research Assistant; built a three-stage pipeline (RTMPose, MotionBERT, HaMeR) that labels co-speech gestures in place of hand annotation, running 97 controlled experiments to choose each stage, and the winning 3D pose representation doubled segmentation F1. Jan–Aug 2025: Amazon, Software Engineer Intern; built an AI agent on LangChain and Bedrock that autonomously generates and debugs build configurations from requirement docs and CDK packages using task orchestration, context engineering, and a knowledge base (RAG), cutting configuration time by 50% and adopted as the team's standard tool; and designed the cost-estimation step in Amazon's internal robot-simulation platform (React, Spring, DynamoDB), back-tested against historical runs and held within 20% error. Sep 2023: Northeastern University, M.S. in Computer Science — the career pivot. 2015–2023: 7 years in product roles in fintech and enterprise SaaS. Sep 2011 – Aug 2015: Peking University, B.S. in Psychology and B.S. in Economics.

🚀 Projects

$ ls -1 ~/projects

sonari/        # real-time voice AI companion · Rust + iOS · 0.95 s median turn
dev-harness/   # guardrails for Claude Code · 11-tool MCP server · 61 tests
gochat-scale/  # Go chat server capacity study · k6 · 7,500 req/s at 224 ms p95

🎙 sonari

Real-Time Voice AI Companion

  • Built and shipped it as an iOS app on a Rust backend deployed to AWS, and piloted it with prospective customers to purchase intent
  • Cut end-to-end turn latency to a 0.95 s median and 1.2 s p95 by restructuring the real-time speech path
  • Built an eval harness of 16 adversarial clips that scores word error rate and eight latency markers to catch regressions

Rust Swift LiveKit AWS

🛡 dev-harness

Guardrails for Claude Code

Built a plugin that enforces three independent controls on a project:

  • Docs-as-code — project documentation is curated, not drifting
  • A ten-stage workflow the agent may request but cannot advance
  • An isolated per-task workspace

Implemented as an 11-tool MCP server with 61 tests across Linux and Windows.

TypeScript MCP

📡 gochat-scale

Scaling an Open-Source Go Chat Server

  • Built a k6 step-ramp load-testing model that measures only each step's steady-state window
  • Used it to place a forked Go chat server's capacity ceiling at 1,500 virtual users sustaining 7,500 requests per second at 224 ms p95

Go Docker k6

Email

Pinned Loading

  1. sonari sonari Public

    Rust

  2. dev-harness dev-harness Public

    JavaScript

  3. gochat-scale gochat-scale Public

    Go

  4. PriceSnap PriceSnap Public

    Go 1 1

  5. KV-Raft KV-Raft Public

    A Kv system based on Raft

    Go

  6. llm-translator llm-translator Public

    JavaScript