I am a backend engineer at Tulips Technologies in Kathmandu. By day I build services and REST APIs in C# and ASP.NET Core on SQL Server, and I integrate generative-AI APIs into enterprise workflows. This site is where I write down what I learn doing that.
What I will write about
- Data pipelines that process a lot of records and have to say how far to trust each one.
- Data quality: cross-checking an AI model's output against other sources before it is stored.
- Observability: logging and metrics with Prometheus and Grafana, so a pipeline can tell you when it is wrong.
- .NET and Python backends: schemas, stored procedures, scraping, and the APIs that serve the results.
Posts will be short notes or longer write-ups, and I will mark which is which.
An example of the tone
Here is the kind of small, concrete thing I mean. When two independent sources agree on a field, accept it; otherwise send it for review.
public enum Verdict { Accept, Review }
public static Verdict Judge(IReadOnlyList<bool> sourcesAgree) =>
sourcesAgree.Count(agrees => agrees) >= 2 ? Verdict.Accept : Verdict.Review;The reason it helps is arithmetic. If each source is wrong independently with probability , then the chance that two of them are wrong together is , which is smaller than whenever . The independence assumption is the part that needs checking, and that is a good topic for a future post.
How the site is built
It is a Next.js site. Posts are MDX files in the repository, validated at build time, with syntax highlighting and math rendered then too. The background animation on the home page is decorative; every post is plain text on a plain page.
You can follow along through the RSS feed.