California · Open to Applied AI / FDE roles

Applied AI · Agent Infrastructure · Product Delivery

I build AI systems that can act, recover, and ship.

Applied AI & Agent Systems Engineer

I turn ambiguous workflows into reliable products — from multi-agent orchestration and tool-use runtimes to context pipelines, evaluation, cloud delivery, and full-stack interfaces.

Explore the work Choose a resume GitHub ↗

Featured systems

Architecture is part of the product.

Four examples of taking an AI workflow beyond a demo: explicit boundaries, governed tools, recovery paths, and interfaces that expose what the system is doing.

Featured · Desktop AI Learning

Trace

Codebase Learning Studio

A desktop-first learning environment that turns unfamiliar repositories into adaptive, source-grounded courses.

Electron React TypeScript Monaco Local Agents
Private preview · GitHub release forthcoming
Trace welcome screen for starting a codebase learning path
Trace adaptive skill tree beside a source-linked lesson
Trace code workspace with a transparent context-budget panel

REPOSITORY INDEX

Maps files, symbols, entry points, languages, Git versions, and source fingerprints locally.

ADAPTIVE ROUTE

Builds diagnostic-driven skill graphs and marks only source-affected mastery stale.

GROUNDED TUTOR

Links Monaco source, local lookup, lessons, and transparent context budgets.

SAFE PRACTICE

Creates isolated Git worktrees with diff inspection and guarded cleanup.

FeaturedTypeScript

AgentFlow

A graph-based orchestration platform for agents that collaborate, use tools, pass quality gates, and produce auditable artifacts.

Private repository · architecture overview

WORKFLOW MODEL

Typed steps, roles, gates, routers, milestones, parallel forks, and human approvals.

MISSION ENGINE

Validation contracts, isolated worktrees, checkpoints, crash recovery, and evidence bundles.

GOVERNANCE

Tool permissions, filesystem and shell policies, edit budgets, approvals, and bounded loops.

OPERATIONS

REST/WebSocket APIs, append-only events, metrics, context compaction, and a real-time React UI.

InfrastructureRust

Agent Bridge

Disposable, sandboxed compute for AI agents on infrastructure you control.

  • Five-verb CLI for run, expose, store, audit, and destroy
  • Local Docker and AWS execution with secrets and storage tiers
  • Cost accounting, retries, diagnostics, bounded audit logs, MCP and SDK surfaces
Private repository · architecture overview
AI Team Builder interface showing generated specialist agents
Product systemNext.js

AI Team Builder

A production multi-agent experience that translates a project brief into specialist roles, tools, workflow, and scoped memory.

prompt caching context compression streaming UI

Open-source lab

Tools for when agents leave the chat box.

WHEN

Agent Hook

Turns cron, webhook, GitHub, and filesystem events into bounded agent runs with capability whitelists, retries, approvals, notifications, and audit history.

Python · SQLite · Webhooks Private repository

CONTEXT

In-Context Search

A Chrome extension for select-to-explain and in-page search, combining surrounding-page context with six AI/search providers across inline and side-panel experiences.

Chrome · JavaScript · LLM APIs GitHub ↗

FEEDBACK

Kaggle Team

Researcher, PM, and Engineer agents improve competition submissions through score-informed rounds, typed tools, parallel strategies, cost tracking, and CV-to-LB feedback.

TypeScript · Bun · Zod Private repository

RENDER

jsfetch

A Playwright-powered CLI, MCP server, and agent skill that renders JavaScript-heavy pages and returns clean text, HTML, or structured JSON.

Python · Playwright · MCP GitHub ↗

Experience

One applied-AI role, two connected workstreams.

The common thread is building reliable inference systems: assembling context, structuring outputs, assessing quality, and delivering results through usable products.

Machine Learning Engineer

SiriusMindshare LLC · San Jose, CA

PRODUCT SYSTEMS

Architected a five-agent product with scoped personas, tools, and memory; engineered cache-aware prompt assembly and four-stage context compression for long-running sessions.

APPLIED RESEARCH

Co-developed an LVLM framework for retail attention-quality assessment, extending the same focus on structured inference and measurable output quality.

Full-Stack Software Developer Intern

1 Thing Against Racism Organization · Remote

Built cross-platform React/React Native/Firebase experiences and Flask/GCP APIs with real-time data, OAuth, encryption, caching, notifications, and AI-assisted content moderation.

Selected publication

Accepted · Forthcoming

From Aisle Video to Shelf Intelligence: A Generative AI and LVLM Framework for Retail Attention Quality Assessment

IEEE 9th International Conference on Computer and Communication Engineering Technology (CCET 2026)

Accepted for publication and presentation at IEEE CCET 2026. The manuscript is forthcoming and is not yet publicly available.

Working set

Systems depth, product range.

Agent & LLM Systems

Orchestration, tool use, context and memory, evaluation, guardrails, prompt caching, structured outputs, MCP

Languages & Product

TypeScript, Python, Rust, JavaScript, React, Next.js, React Native, FastAPI, Node.js

Cloud & Runtime

AWS, GCP, Docker, Lambda, SQS, DynamoDB, CloudWatch, CDK, WebSocket systems

Data & Retrieval

PostgreSQL, SQLite, Redis, Elasticsearch, Firebase, RAG pipelines, browser-rendered extraction

2022 — 2025

MS Computer Science · 4.0 GPA

Northeastern University · San Jose, CA

2015 — 2019

BA Quantitative Economics · 3.5 GPA

University of California, Irvine

Let's build

Need an AI system that survives contact with the real workflow?

I'm interested in Applied AI, AI/LLM Systems, and Forward Deployed Engineering roles where architecture and delivery matter equally.

Email June +1-213-477-3071 Choose Resume