Backend & APIs
Application logic and integrations
- PHP 8
- Laravel
- Python
- REST & JWT
- Networks & Linux since 2011
- Web work since 2014
- Worked with teams worldwide
Seyed Reza Bazyar
I start at the root of a problem, then build only what it needs.
At sixteen, I set up home networks, learned hardware, installed Linux, and wrote small scripts by hand. I moved into PHP web work in 2014, building apps, APIs, and database-backed features. I still read the code, query plans, and logs myself. I build AI agents and bots to automate repetitive work and reduce the time and cost it takes.
Outside of code, I love books.
- Networks & Linux
- Since 2011
- PHP web work
- Since 2014
- Everyday tools
- Laravel · Nuxt · Elasticsearch · Python
- [email protected]
- 2011Networks, Linux, and hand-written code
- 2014PHP and web systems
- NowAgents and automation
Terminal response
I started with home networks, hardware, and Linux at sixteen. I wrote small scripts by hand, then moved into PHP web work in 2014.
Skills & tools
The parts of a web system I work with day to day.
Data & search
Storage, read paths, and retrieval
- MySQL / PostgreSQL
- MongoDB
- Elasticsearch
- Redis
Infrastructure
The services and networks beneath the app
- Linux & systemd
- Nginx & Apache
- Docker
- Networks, DNS & TLS
Web & frontend
Rendered pages and browser-side code
- Nuxt (SSR)
- TypeScript
- Tailwind CSS
- WordPress & WooCommerce
AI & automation
Repeatable work with human review
- Cooperating agents
- LLM pipelines
- Telegram bot
- Data collection & processing
Delivery & operations
Changes that teams can follow
- Git & CI
- Docs & task writing
- Monitoring & alerts
- Security review
Past work
A few kinds of engineering problems I have worked through.
- Search
Multilingual search and retrieval
Persian and multilingual search returned uneven results, and the index lagged behind its source. I normalized text and ranking, then kept Elasticsearch aligned with the primary database.
- Data
Read models for database-backed systems
The relational database served every read pattern. I split out a read model, used keyset pagination, and tuned indexes from query plans.
- Performance
Speed and resource cost
Hot paths raised latency and resource use. I traced the bottleneck, separated cache from queues, set service budgets, and used Octane where it fit.
- Reliability
Health checks, alerts, and recovery
Noisy alerts and risky deploys slowed recovery. I added a light health endpoint, central alerts, rollback steps, and blame-free postmortems.
- Networks
Edge behavior and international routes
Traffic varied by route and crawler. I tuned proxy rules and rate limits, measured CDN effects, and traced international issues through MTU and PMTUD.
- Security
API access and data exposure
API responses and downloads had unclear access boundaries. I reviewed data exposure, used signed links and per-user checks, and documented vulnerability disclosure.
- Data work
Repeatable collection and processing
Public-source data arrived in mixed formats. I used rate-limited crawlers and queues, normalized records into SQL, and made the pipeline repeatable.
- Agents
Agent workflows with review points
Agent workflows needed cost and approval boundaries. I compared batch and live model runs, used multi-agent steps where they fit, and kept code review and human approval in the loop.
How I work
Small steps, written down as we go.
Logs, query plans, and curl first; then a small, reversible change, documented in the same commit.
Check the evidence first
Read the logs, query plan, or request before changing code.
Keep changes reviewable
Make a focused change that a teammate can measure and roll back.
Write the task and decision down
Put the task and its reason in writing; keep the notes with the commit.
Treat cost as a requirement
Choose the smallest tool that meets the need, then compare its running cost.
Automate repeat work; keep human approval
Agents can handle repeatable steps. People keep the judgment and sign-off.
Leave a usable handoff
Include a runbook, monitoring, alerts, and notes the team can use without me.
Problems I work on
A few examples of where I can help.
If
Your team repeats the same slow task
I do
I map the steps, automate the repeatable work, and keep review and traceability in place.
If
Search or data is hard to trust
I do
I shape the search index, read model, or processing pipeline around how the data is used.
If
A backend or API is getting in the way
I do
I trace the request, then work on the Laravel app, queue, or integration that needs attention.
If
The system is slow or costly to run
I do
I find the measured bottleneck and reduce wasted work or resource use.
If
An unreliable system keeps you up
I do
I work through reliability, monitoring, recovery, and access controls.
If
A WordPress store needs custom work
I do
I build the plugin or automation missing from the publishing and commerce flow.
If your problem isn’t on this list, tell me what’s going on.
Working with a team
A clear scope and a handoff the team can keep using.
How we get started
Describe the problem
Tell me the outcome, constraints, and budget you have in mind.
Get written options
I’ll outline two or three ways forward, with trade-offs and costs.
Start with a small working version
Build and review a runnable slice before expanding the scope.
Hand over the operation
Leave documentation, monitoring, alerts, and next steps with the team.
Ways to work together
- Architecture and code review
- Stabilization and incident work
- Agent and automation sprint
- Cost and performance review
- Build from idea to production
- Work inside your team
What a handoff can include
- Written decisions and documentation
- Monitoring and alerts
- A system cost report
- A list of next steps
Questions teams often ask
What kinds of problems do you take on?
My work spans backends and APIs, search and data, performance and cost, reliability and security, and task automation. If your issue is not on that list, describe it in the form.
Where do AI agents help, and where do they not?
They fit repeatable work with clear inputs and results someone can review. Decisions with real consequences and final approval stay with a person. If checking the output costs more than doing the task, an agent is the wrong fit.
How do you keep costs down?
Start with logs, query plans, and resource use. Fix the measured bottleneck, choose the smallest tool that fits, and check the result against its running cost.
Can you work with our existing team?
Yes. We can agree on the task format, review process, access, and handoff before work begins.
How do you hand over code and documentation?
Changes stay reviewable in Git. The handoff can include runbooks, monitoring, alerts, written decisions, and next steps.
How do we start?
Send the problem, the outcome you need, constraints, and a budget range. I’ll reply with written options and the trade-offs to discuss.
Tell me what you’re working on
I work remotely with teams worldwide. Send the goal, constraints, current behavior, and budget range; we can start with a written assessment.
- [email protected]
- Working arrangement
- Remote · worldwide