I owned it from idea to production
I worked through the API boundaries, strategy logic, browser pools, and deployment instead of stopping at a prototype.
Backend, cloud, and data
Building the systemsbehind the interface.
I’m Parash. I build APIs, data pipelines, and cloud infrastructure that keep working when the traffic, data, and edge cases pile up.
4+ years building microservices, real-time pipelines, and AWS infrastructure · Based in Kathmandu, Nepal
Parash SubediBackend Engineer at Grepsr
4+Years building backend systems
MillionsEvents handled every day
2025Grepsr Employee of the Year
AWSSolutions Architect, Associate
01
A system I built
Production / Built solo
Grepsr · ScrapeSwitch
A scraping platform for pages that do not play nice.
I built ScrapeSwitch solo for Grepsr. A Rust API decides how each request should run, remembers the session, resolves a proxy, and hands browser-heavy work to a Python fetcher pool.
System mapRequest → result
Incoming requestValidate · normalize
OrchestratorRust · Axum · Tokio
Redis memorySession read / write
Strategy decisionSession · proxy · strategy
Static fetchcurl-impersonate · HTTP
Fetcher poolPython · FastAPI · Playwright · patchright
ResultAdaptive output
Static-compatible? HTTP. Browser required? Fetcher pool.
Validate and normalize the incoming request before choosing a fetch strategy.
Fast when possible, browser when needed
Some pages only need a static request. Others need a full browser. ScrapeSwitch hides that choice behind one API.
Rust decides, Python fetches
Rust keeps the control path lean. Python workers take on the slower browser work.
02
Day-to-day impact
The API is only part of the job.
At Grepsr, I work across backend, data, and infrastructure. That makes it easier to follow a production problem to the real cause instead of passing it between teams.
Current roleBackend Engineer
GrepsrFeb 2023 to present
Services that are easy to run
I build Python, Go, and gRPC services, then carry them through EKS, Helm, and CI/CD so the team can ship and run them without guesswork.
- A programming language I use to build services and process data.
- A programming language suited to services doing many jobs at once.
- A defined way for separate programs to ask each other for work and exchange data.
- Amazon’s managed Kubernetes service for running applications across computers.
- Packages the settings needed to install or update an application on Kubernetes.
Millions of events, every day
I work on Java, Flink, Fluentd, and Kinesis pipelines that process millions of events a day, with Temporal and AWS services keeping the workflows moving.
- A programming language used for long-running services and large data workloads.
- Apache Flink checks and transforms streams of data as they arrive.
- An Amazon service that gathers streams of data for other systems to process.
- Keeps multi-step jobs moving, including retrying failed steps after interruptions.
- An Amazon service that runs a piece of code when triggered, without a dedicated server to manage.
The work around the endpoint
Auth, permissions, storage, observability, and database tuning are part of the job too. That work helped bring high-traffic API latency down from seconds to milliseconds.
- Checks who a person is when they sign in.
- Checks what a signed-in person is allowed to see or do.
- A database that stores related records in tables and checks rules about that data.
- Keeps frequently needed data close at hand so it can be read quickly.
- An Amazon service for asking questions of data stored in S3 using database queries.
AI that solves a real problem
I built tools that generate data quality rules and pull CSS selectors and XPaths from crawled pages. They run inside real pipelines, not as side demos.
- Checks whether data is complete, correct, and usable.
- Amazon’s storage service for files, exports, and other saved data.
- An Amazon service that runs a piece of code when triggered, without a dedicated server to manage.
- Keeps multi-step jobs moving, including retrying failed steps after interruptions.
03
Open source
Fixes I sent upstream
If I can fix it upstream, I do.
These are small, specific changes, but they now help everyone using the project.
01
Python
fivetran/
I fixed two issues I ran into while working with Great Expectations, then sent the changes upstream.
30k+ GitHub stars02
Swift / TypeScript
TableProApp/
I traced a DeepSeek V4 streaming failure in TablePro and fixed the chat path that was returning 400 errors.
4.8k+ GitHub stars03
Python
D4Vinci/
I helped fix a locale-dependent bug in Scrapling’s Cloudflare challenge handling.
Released in v0.4.1504
Recognition
Grepsr’s 2025 Employee of the Year.
Grepsr named me its top-performing employee of the year.
Hackathon Champion: CrawlGuard
I brought a cross-functional team together and we built CrawlGuard, a real-time data quality system, in 48 hours. It won the internal hackathon.
AWS Certified Solutions Architect, Associate
I earned the certification in 2025 and use the same ideas in production every day.
05
About and toolkit
Parash SubediParash Subedi · Kathmandu, Nepal
I like clear boundaries, plain code, and quiet on-call shifts.
I’m happiest when I can trace a request from the API through the data and into the infrastructure. I care less about clever code and more about systems that are easy to understand, change, and run.
EducationBachelor of Computer Science
Kathmandu University · 2018 to 2023
Select a technology for its definition.
Languages
- A programming language I use to build services and process data.
- A programming language suited to services doing many jobs at once.
- A programming language used for long-running services and large data workloads.
- A programming language that does the work on the computer running a website.
- A programming language designed for speed and memory safety.
Backend and data
- A Python tool for letting other software send requests and receive answers.
- A defined way for separate programs to ask each other for work and exchange data.
- A Rust tool for receiving web requests and sending responses.
- A Rust tool that lets other tasks progress while one waits for data.
- A system that checks and transforms streams of data as they arrive.
- Keeps multi-step jobs moving, including retrying failed steps after interruptions.
- Controls a real browser to open pages, click, and read what appears.
Cloud and delivery
- Amazon Web Services provides computing, storage, and other services on demand.
- Bundles an application with what it needs to run consistently on different machines.
- Starts and manages packaged applications across a group of computers.
- Packages the settings needed to install or update an application on Kubernetes.
- Runs checks and release steps automatically when code changes.
Storage and access
- A database that stores related records in tables and checks rules about that data.
- A database for storing and querying related data in tables.
- A database that stores records as flexible documents rather than fixed table rows.
- Keeps frequently needed data close at hand so it can be read quickly.
- Amazon’s storage service for files, exports, and other saved data.
- Checks who a person is when they sign in.
- Checks what a signed-in person is allowed to see or do.