Engineering Highlights
A selection of things I've built, shipped, or contributed to — across engineering, writing, and developer advocacy.
AI Workflows Execution Sandbox
Built an isolated Python execution engine for an AI-assisted workflow builder. Each sandbox ran on a persistent Python kernel — similar to Jupyter or Marimo — preserving execution state between runs, with stdout/stderr streamed back to the frontend. A pluggable FileSystem abstraction (local today, S3/memory ready) gave each sandbox a temp working tree split into source (read-only inputs copied in at startup) and target (execution output), with every operation validated against the sandbox root.
Python, FastAPI, asyncio, subprocess, WebSocket
Python Runtime Infrastructure
Built a server that accepted Python scripts via API, resolved and downloaded their dependencies, and packaged everything into a self-contained binary using PEX. Each task ran inside a Firecracker microVM for fast, lightweight sandboxing — giving users the ability to execute arbitrary code safely with full isolation. Tasks were queued through Redis, with worker servers pulling and executing jobs concurrently.
Python, PEX, Firecracker, Redis, REST API
Custom Formula Editor
Built an Excel-like formula editor from scratch — including a custom parser that performed lexical and syntactic analysis, produced an AST, and applied transformations to evaluate expressions. Supported user-defined formulas for computed columns and data manipulation within the platform.
TypeScript, AST, Parser, Lexer
ClickHouse Sharding Configuration
Designed and configured sharding for ClickHouse to handle the platform's analytical workloads at scale — setting up distributed tables, shard topology, and replica configuration for high-throughput data ingestion and querying.
ClickHouse, distributed systems
AI Interview Engine
Built an AI-powered pre-screening interview engine. Given a position and skill requirements, the system used LLMs to auto-generate a structured interview plan — a tree of questions with scoring rubrics and follow-up logic. During the interview, an LLM harness orchestrated multiple calls to conduct the conversation, evaluate candidate responses, and decide on follow-up questions in real time.
Python, LLM orchestration, structured generation
Monorepo Build Automation
Designed and implemented schematics-style build automation for an Nx-based monorepo. The generators handled the full scaffolding lifecycle: templating files, applying AST rewrites to update routing, module, and barrel files, and patching shared configs like nx.json and tsconfig.base.json. Reduced the cost of spinning up new features and enforced architectural conventions at generation time.
Nx, TypeScript, AST, code generation, monorepo tooling
Community & Content
Built the largest Angular publication from scratch. Wrote 100+ in-depth articles, many ranking in the top 3 of Google search results. Grew the publication to 1M+ monthly views and 16k+ followers. Recruited and edited a community of 20+ writers, built an internal Slack group for reviews and collaboration. The publication became the go-to resource for advanced Angular content.
Content strategy, SEO, community building, developer advocacy