01 / Founding engineer fit
Ship the first serious version of an AI-native product surface.
Best for early teams that need someone who can move from ambiguous workflow to product, UX, validation, docs, and launchable implementation.
AI-native developer tooling / agent workflow builder
I build agent-native developer tools that turn vague workflows into reliable systems.
I am an applied AI product engineer focused on agent workflows, developer tools, technical documentation, memory, validation, and the product surfaces that make AI systems useful to teams.
Why this is credible
A repo is not a company. It is proof of a scarce operating capability.
A repo is not a company. It is proof of a scarce operating capability.
The strongest signal in this body of work is not prompt novelty. It is the ability to design constraints, verification, export, reuse, and developer-facing workflow around AI systems.
Public OSS traction from developer-facing AI tooling
Can turn vague workflows into constrained, reusable systems
Comfortable across product framing, implementation, validation, docs, and distribution
Strong fit for agent platforms, AI IDEs, devtools, internal copilots, and applied AI workflow startups
How I can enter a team
Start with the missing module, not with a vague cofounder conversation.
02 / Design partner sprint
Turn a fragile internal AI workflow into a repeatable system.
A two-week build sprint around agent tooling, developer workflows, internal copilots, technical diagrams, memory, or docs automation.
03 / Contract-to-founder
Start with a concrete module, then test operating fit.
A low-drama way to discover whether the collaboration should become deeper: build something real, measure judgment, then decide.
Live signal / X
Current notes from the edge of agentic devtools.
Synced from the official X API into the site content ledger. The public site stays English-first for now; Chinese routes are temporarily redirected.
X sync is ready; waiting for the first production ingest.
Once the cron has the X bearer token in Vercel, this panel will render the latest @teach_fireworks notes automatically.
Built in public
A small portfolio of AI-native workflow assets.
01 / Python / 5.4k stars
fireworks-tech-graph
A project that treats architecture diagrams not as cosmetic add-ons, but as durable explanatory surfaces for AI systems work.
02 / Python / 91 stars
fireworks-skill-memory
A memory system for coding agents that values persistence, reuse, and editorial recall over mere accumulation.
03 / Python / 8 stars
fireworks-sessions-saver
A checkpointing layer for coding sessions, built to survive interruption, context loss, and handoff without pretending memory never breaks.
04 / TypeScript / 2 stars
paper-pal
A paper-reading tool that frames comprehension as an active workflow of margin notes, synthesis, and return visits.
Outreach thesis
I am looking for teams where this capability is currently missing.
The right first conversation is specific: your product has an AI workflow that needs to become reliable, inspectable, documented, and useful enough for developers or operators to adopt. I can help build that surface, then we can decide whether the relationship should continue as contract, founding engineer, or deeper founder fit.