Field notesKathmandu, Nepal · UTC+5:45

I keep clouds up,
costs down, and
pagers quiet.

I am Prasanga Kadel, a senior DevOps engineer. At Genese Solution, an AWS Premier Partner, I look after the AWS estate: architecture, guardrails, the bill, and the on-call rotation that keeps all of it honest.

Start a conversation Read the work

01Selected work

Three infrastructure problems, and what changed.

Each entry follows the same shape: context, constraint, approach, outcome. Details are generalized to respect client confidentiality.

Case 01 Genese Solution Jul 2023 — Present EC2, VPC, IAM, Organizations, Control Tower

AWS at scale, without click-ops

More than 40 accounts, each with its own drift, access paths, and audit evidence. The work was to make them behave like one system without freezing the teams that use them.

Reference pattern: a load balancer fans out to application instances in two availability zones, security groups chain from load balancer to app tier to database, and RDS runs with a primary and a replica. CLIENT REQUESTS ALB App · AZ-a security group: app App · AZ-b security group: app RDS primary security group: db RDS replica separate AZ SECURITY GROUPS CHAINED · ALB → APP → DB
  • ALB → app tier in two availability zones → RDS primary and replica.
  • Security groups chained: ALB → app → database.
Simplified reference pattern applied per account. Account IDs, addresses, and customer details omitted.

Constraint

Every account stays up and stays secure, and can prove it when an auditor asks. No snowflakes, no bespoke exceptions.

Approach

Terraform and Ansible for repeatable provisioning. Control Tower guardrails for the boundaries. AWS Config and FTR reviews to produce evidence instead of assurances.

OutcomeOne operating model across 40+ accounts, and 100% compliance with AWS Premier Partner requirements.

Case 02 FinOps Cost Explorer, rightsizing, reserved capacity

The bill, read properly

Spend had grown the way it does when nobody owns the number: a little each month, always explainable, never examined.

Constraint

Cut the internal account bill without cutting the redundancy that exists for a reason. No heroics, no guessing.

Approach

Read usage data before touching anything. Rightsize what was oversized, reserve what was steady, and check multi-cloud estimates against measured usage.

OutcomeInternal spend down 20%. Budgets now start from measured usage instead of last quarter's guess.

Method

  1. 01Read usage before changing it.
  2. 02Rightsize what was oversized.
  3. 03Reserve capacity for steady load.
  4. 04Re-baseline budgets on measured usage.

Relative internal spend index. Absolute figures are internal.

Case 03 CI/CD + SRE GitHub Actions, Jenkins, CloudWatch, Python

Deploys that stopped needing courage

Releases ran on manual steps and whoever happened to be awake. The goal was never zero humans. It was zero waiting.

Pipeline: git push triggers GitHub Actions for build and test, then deploy. A dashed feedback path leads from deploy to CloudWatch, then runbook, then post-incident RCA. COMMIT TO PRODUCTION git push GitHub Actions build + test deploy FEEDBACK OBSERVE AND RESPOND CloudWatch runbook RCA
  • git push → GitHub Actions (build + test) → deploy.
  • CloudWatch → runbook → post-incident RCA.
Before: manual steps. After: a repeatable path to production, and a feedback loop for the incidents that still happen.

Constraint

Ship faster without shipping fear. Stop re-fighting the same incident every month.

Approach

Pipelines for the repeatable parts. CloudWatch with Python for bottleneck detection. Runbooks, playbooks, and RCAs for the parts that still need judgement.

Outcome

Cloud-native deploys 40% faster, manual intervention down 25%. The on-call rotation still exists; it is just quieter.

02Results

What the work adds up to.

Measured against the operating baseline before each change. Relative changes are reported; absolute figures stay internal.

40+

AWS accounts under one operating model

Case 01

20%

Lower internal infrastructure spend

Case 02

40%

Faster cloud-native deployments

Case 03

25%

Less manual intervention

Case 03

Baselines and measurement windows vary by project. No figures are projected or model-generated.

03Experience

Eight years, three disciplines.

QA and GIS came first; the move into DevOps started in 2023. The earlier roles shaped how I read systems before changing them.

  1. Jul 2023 — presentKathmandu

    DevOps Engineer Genese Solution

    AWS Premier Partner

    • AWS operations across 40+ internal and customer accounts (EC2, S3, VPC, IAM).
    • Lead on-call rotations and incident response: runbooks, playbooks, post-incident RCAs.
    • AI-assisted operations with Claude Code and KIRO; MAP / POC / BOX funding tracking.
  2. Jan 2023 — Jun 2023Bagmati

    DevOps Apprentice Coderush

    • CI/CD with GitHub Actions and Jenkins; Ansible automation.
    • Delivered 40% faster cloud-native deployments.
  3. Feb 2018 — Nov 2021Nepal

    GIS Consultant / Analyst EERC, CCRC & New Planet Engineering

    • Automated geospatial workflows with Python: 30% faster data processing.
  4. Nov 2016 — Nov 2017Lalitpur

    QA Engineer GrowByData

    • Functional and regression testing across the SDLC; standardized QA protocols.

04Capabilities

A working index, not a badge wall.

Everything listed here is in use or was used in production. The terminal below answers to the same list.

AWS & Cloud
EC2, VPC, S3, RDS, CloudFront, Route 53, Migration Hub, Organizations, Control Tower, Cost Explorer
Infrastructure as Code
Terraform, Ansible, AWS CodePipeline
CI/CD & Automation
GitHub Actions, Jenkins
Containers
Docker, Kubernetes, ECS, Fargate
Observability
CloudWatch, Grafana, Prometheus, Nagios
Security & Governance
IAM, AWS Config, WAF, Systems Manager, FTR
Scripting & Linux
Python 3, Bash, Linux
AI-assisted Operations
Claude Code, KIRO

Certifications

  • AWS Certified Solutions Architect — Associate
  • AWS Certified CloudOps Engineer — Associate
  • AWS Certified Security — Specialty
  • AWS Certified AI Practitioner

Verify on Credly

BSc CSIT, Tribhuvan University, 2012–2016.

05Terminal

The field terminal.

Optional, and honest: it restates the record above rather than watching any live system. Type help, or use a hint. History and tab completion work.

help
list every command
projects
the case studies, condensed
experience
the same four roles
contact
email and profiles
field-terminal — ~/portfolio
~/portfolio $
Try these:

06Contact

Tell me what is actually hurting.

If the bill keeps climbing, deploys feel risky, or the pager owns your nights, describe the problem. You will get a straight scope and an honest answer about whether I can help.

Request a consultation
Email widpra@gmail.com
LinkedIn linkedin.com/in/widpra
Credly credly.com/users/prasanga-kadel
Location Kathmandu, Nepal · UTC+5:45
Booking 30-minute consultation
Copied