Backend
Python, ASGI-style APIs, application boundaries, validation, data flows, background work, and operational behavior.
About Peter
I learned software after years of building things that could not be vague.
My background is construction, operations, sales, and independent software systems. The common thread is accountability when the plans, the field conditions, and the real failure do not agree.
01 / Path
Carpentry and field construction
I read plans, sequenced work, managed tolerances, and solved undocumented conditions where a polished explanation did not matter if the structure failed.
Rental and maintenance responsibility
I managed interdependent work, people, contracts, and failures where there was no separate department to absorb the problem.
Turning confusion into an actionable path
Inside sales taught me to listen for the actual constraint, explain the decision without unnecessary machinery, and ask directly for the next action.
Software and AI systems
I moved into Python, Postgres, APIs, deployment, retrieval, automation, and web systems by building complete projects and auditing the parts that failed.
02 / How I work
I do not need the first theory to be right. I need the evidence trail to survive being wrong.
- Start from behavior
Reproduce the thing that failed before changing the architecture around it.
- Control scope
Protect working behavior and refuse the temptation to rebuild adjacent systems merely because they are interesting.
- Test the claim
A test matters when it would fail if the behavior broke. Green output without that relationship is theater.
- Document limits
I would rather state what is not proven than bury a weakness under confident language.
- Complete the handoff
The work includes the production state, relevant account ownership, operating steps, and what should be checked next time.
03 / Technical range
Data
Postgres, pgvector, Neon, Cloudflare D1, schemas, migrations, queries, persistence, and event records.
Web and deployment
Static and JavaScript sites, Cloudflare, Vercel, Fly.io, GitHub, DNS, domains, email integrations, and production verification.
AI systems
Memory state, retrieval, embeddings, evals, token accounting, governance constraints, ingestion, and failure analysis.
04 / AI tooling
I use AI tools. I do not transfer responsibility to them.
AI helps me inspect, compare, generate, challenge, and accelerate work. It can also produce convincing garbage, expand scope, repeat fashionable patterns, and claim completion without evidence.
My responsibility is to define the contract, inspect the output, run the tests, verify the production behavior, and state what remains uncertain. The final work is mine.
05 / Work with me
Send the problem before the job title.
For immediate commercial work, the clearest entry is website technical rescue. For backend, AI infrastructure, automation, internal tooling, or a role where broad systems diagnosis matters, email peter@peterbaumgardner.com.