req_7f3a9cGET /pending

Raghav Kattel

I build systems that turn complex problems into working products.

Senior Software Engineer in Kathmandu, working across distributed systems, AI applications and blockchain infrastructure — and the product decisions that connect them. Scroll to follow one request through the kind of systems I design.

Explore selected work All projects

Scroll to send the request

A simulated request. Nothing on this page calls real infrastructure.

01 · Ingress

A system is more than one process.

The request reaches a load balancer. Instead of one server doing everything, work is routed across several identical service instances, so no single process becomes the bottleneck — or the single point of failure.

REQUEST RECEIVED
ROUTING
SERVICE EXECUTION

Ingress: Terminates TLS and admits traffic into the cluster.
Load balancer: Spreads requests across the instances that are currently healthy.
Service instances: Identical, stateless replicas — any of them can serve the request.

02 · Failure

Reliability shows up when something breaks.

One instance stops answering. Health checks mark it unavailable, the router stops sending it work, and the in-flight call is retried on a healthy instance while a replacement is scheduled. Nothing magical — redundancy, health checks and retries, designed in from the start.

HEALTHY
DEGRADED
UNAVAILABLE
RECOVERED

Health check failed: Missed probes mark the instance unavailable and routing stops sending it traffic.
Retry: The failed call is retried once on a healthy instance. Idempotent handlers make the retry safe.
Replacement scheduled: The orchestrator starts a new instance to restore capacity.
Merged 3/3: All three calls return and the response is assembled.

AIOps and Self-Healing Infrastructure: Myth vs Reality for DevOps Engineersnote

03 · Intelligence

Some requests need more than deterministic code.

When a request needs judgement, it moves through an AI pipeline: the input is embedded, relevant knowledge is retrieved, a context window is assembled, and a model streams back a response. The engineering lives in retrieval, grounding and evaluation — not in the magic.

INPUT
RETRIEVAL
CONTEXT
INFERENCE
STREAM

Embedding: The query becomes a vector so it can be compared by meaning, not keywords.
Retrieval: The nearest passages are pulled from a vector index.
Context assembly: Retrieved passages are ranked, trimmed and kept with their sources.
Inference: The model reasons over the assembled context.
Streaming: Tokens stream back as they are generated, so progress is visible immediately.

04 · Transaction

Coordinating actions beyond a single database.

Some requests must be verifiable by parties who don’t share a server. The request is signed and becomes a transaction: submitted, validated, included in a block and confirmed. Account abstraction lets an application sponsor and batch those steps, so the person on the other end never has to manage gas or keys.

SIGN
SUBMIT
VALIDATE
INCLUDE
CONFIRM

Submitted: With ERC-4337, a UserOperation goes to a bundler rather than straight to the mempool.
Validated: Signatures, nonces and paymaster rules are checked before inclusion.
Confirmed: Each block built on top adds a confirmation. Until then it is only a request.

05 · Impact

The system matters because of what it makes possible.

Infrastructure is only interesting when it reaches people. At Rumsan I worked on Rahat, an open-source platform for humanitarian cash and voucher assistance, where a confirmed transaction meant support could reach families faster.

CONFIRMED
DISBURSED
DELIVERED

06 · The work

A request is meaningful when it produces something useful.

Each layer of that journey maps to things I have built. A few of them:

All projects →

07 · Response

req_7f3a9c Request completed

200 OK

The request returns where it started.

I design and build systems end to end — from infrastructure and data models to the AI and product decisions on top. Explore the work, read the engineering notes, or get in touch.

Get in touch All projects

Connection available