Amodh Dhakal

amodhakal@gmail.com GitHub LinkedIn

Education

North Carolina State University

B.S. Computer Science

Aug 2023 - May 2025 · Raleigh, NC · GPA 3.96 / 4.00

Courses

Software Engineering
Data Structures & Algorithms
Operating Systems
Network Security
Discrete Mathematics
C and Software Tools
Automata, Grammars, and Computation
Linear Algebra
Probability and Statistics
Computer Graphics
Animal Computer Interaction
Human Centered Security
Building Game AI
Computational Visual Narrative

Skills

Languages

Python
JavaScript
TypeScript
Java
C#
C
C++
SQL
HTML
CSS
GLSL

Frameworks

.NET
ASP.NET
Entity Framework
Flask
Spring Boot
React
Next.js
Angular
Hibernate
Vue
LangChain
LangGraph
Bun
Node.js
ShadCN
TailwindCSS
PySpark
OpenGL

Tools

AWS
Azure
Docker
GitHub Actions
SQL Server
PostgreSQL
Kafka
RabbitMQ
Terraform
Redis
OpenTelemetry
Jira
Grafana
Prometheus
CMake

Experience

Software Engineer @ Game2Learn Lab

Jan 2025 - Apr 2025

3D medical training platform built with a 5-person agile team and medical faculty. Cut build times 78% down to 28 seconds, shipped the core UI with a live-synced 2D ward map, and held 85%+ test coverage across iOS and Android.

Vue
Vite
Vitest
Fastify
GitHub Actions

Developed the Vue.js interface for a healthcare training simulation, covering ward maps, patient charting, and branching conversations, with workflows validated by medical faculty each week in Agile sprints.
Reached 85%+ test coverage across store modules and views with Vitest and GitHub Actions, catching regressions before production deployment and keeping weekly releases to medical faculty stable.
Translated ambiguous medical faculty requirements into user stories, acceptance criteria, and Jira tasks for a Scrum team on a Kanban board, verifying each through system tests and weekly demos.
Reproduced defects reported by medical faculty, traced root causes, and coordinated fixes with engineers across weekly sprints, shortening the path from bug report to verified fix before faculty demos.
Reduced Vite build time by 78%, to 28 seconds, by manually splitting bundles and pre-bundling dependencies, shortening the local development feedback loop for engineers shipping weekly simulation changes.

Projects

MariaDB

[Source]

Open-source contribution to the MariaDB server engine, hardening reliability and fixing runtime crashes.

C++
SQL
CMake

Contributed C++ fixes to MariaDB’s open-source database server engine across query execution and storage code, with regression tests and changes submitted for public review by core maintainers.
Add check-constraint validation to the InnoDB storage engine, rejecting invalid input that could otherwise trigger a runtime crash, with regression tests preventing the failure from returning.
Fix resource cleanup on error paths in INSERT and REPLACE SQL queries so failed statements no longer leave the server in an inconsistent state, hardening database reliability under invalid input.
Correct inaccurate REPAIR TABLE progress reporting by tracing status updates from the command-line interface into the storage engine, so users see accurate progress during long-running repairs.

QuadroURL

[Source]

Distributed URL shortener on AWS that sustains 12,400+ req/s at 200 ms p95 under load.

Python
Flask
AWS
Terraform
Kafka
Docker
PostgreSQL
Redis
Prometheus
Grafana
Locust
Playwright
Pytest

Build a full-stack, distributed URL shortener on AWS with a Python and Flask backend, Kafka-based asynchronous event processing, and a React/TypeScript frontend for redirects and analytics.
Load-test the service with Locust at 12,400+ requests per second and 200ms p95 latency, meeting targets of under 500ms p95, under 1 second p99, and under 1% errors via cache, index, and connection-pool tuning.
Maintain 80%+ test coverage across backend and frontend code with Pytest, Vitest, and Playwright end-to-end tests, with checks automatically enforced on every code change by GitHub Actions CI.
Improve redirect read performance with an in-process L1 cache and a distributed Redis L2 cache, using request coalescing and jittered TTLs to prevent cache stampedes under heavy redirect traffic.

Linterra

[Source]

3D voxel game engine in C++ and OpenGL with infinite procedural worlds and GPU terrain generation.

C++
OpenGL
GLSL
doctest
CMake
GitHub Actions

Developed a from-scratch 3D voxel engine using C++ and OpenGL, implementing underlying systems for infinite, procedurally generated block-based worlds.
Designed a GPU-accelerated terrain pipeline using compute shaders and Shader Storage Buffer Objects (SSBOs) for parallelized FastNoise evaluation, achieving real-time heightmap generation.
Engineered a high-performance rendering architecture supprting 60+ FPS with 64 chunks with an IRenderer abstraction layer and an OpenGL backend, utilizing 7x vertex compression to store vertex data in 4 bytes and a texture array for efficient rendering.
Implemented a robust multithreaded architecture with a persistent worker thread pool for background chunk generation and meshing, incorporating thread safety and RAII resource management to prevent data races and leaks.
Established a comprehensive testing and CI/CD pipeline, including headless unit tests with doctest and a custom 'smoke test' that drives the GL path with an offscreen context in CI on Ubuntu and macOS.

Policy Manager

[Source]

Full-stack insurance policy administration system handling 13,600+ req/s on Azure Container Apps.

Azure
C#
.NET
ASP.NET
Entity Framework
Angular
SQL Server
RabbitMQ
MassTransit
OpenTelemetry
Polly
xUnit
Selenium

Built a full-stack insurance policy administration system with ASP.NET Core, Entity Framework Core, and Angular covering policyholders, policies, claims, and reporting, persisted in SQL Server.
Achieved 90%+ test coverage with xUnit, Vitest, and Selenium, and built GitHub Actions CI jobs for coverage, containerized integration tests, and EF model-drift checks so failures surface at the correct stage.
Deployed six instances on Azure Container Apps behind a load balancer, sustaining about 13,600 requests per second, with OpenTelemetry tracing and Polly resilience pipelines for fault tolerance.
Implemented a transactional outbox so each domain update and its event commit together, publishing through MassTransit and RabbitMQ with crash recovery, bounded backoff, and dead-letter handling.
Secured policyholder data with AES-GCM encryption, per-holder access auditing, and per-endpoint JWT role checks, plus rate limiting before token validation and request-body limits before model binding.

BrewLogic

Cafe ordering platform with JWT auth, role-based access, and real-time order tracking over WebSockets.

Java
JavaScript
Spring Boot
React
AWS
MySQL
Hibernate
Maven
JUnit

Built and deployed a cafe ordering platform on AWS, using Java Spring Boot and Hibernate for orders, inventory, and menus, with Spring Security JWT authentication and role-based access control.
Automated CI/CD with GitHub Actions and achieved 90%+ test coverage across services and controllers using JUnit, catching regressions before each deployment of the AWS-hosted ordering platform.
Developed a React and JavaScript frontend for the cafe ordering platform, using WebSockets to show order and inventory changes in real time, keeping staff views current without manual refreshes.

uBlockAI

[Source]

Chrome extension detecting AI-generated misinformation in social media feeds in real time.

Python
JavaScript
Flask
LangChain
LangGraph
Tesseract OCR
OpenAI GPT

Engineered a browser extension using Python (Flask) and JavaScript to detect AI-generated misinformation in real-time social media feeds.
Integrated LangChain and LangGraph to orchestrate multiple AI models and OCR (Tesseract) for comprehensive analysis of text and images within posts.
Implemented a multi-stage detection pipeline, including prompt engineering with OpenAI GPT and visual analysis, to identify stylistic and factual inconsistencies indicative of synthetic media.
Developed a seamless user interface for flagging detected misinformation directly within the browser, providing users with real-time credibility assessments of viewed content.

ChartGuard

[Source]

Automated PySpark data validation suite ensuring Synthea patient record integrity in Delta Lake.

Python
PySpark
Spark SQL
Pytest
GitHub Actions

Built a PySpark pipeline that loads Synthea synthetic patient records into Delta Lake tables and derives per-patient encounter summaries with Spark SQL.
Wrote an automated data-validation suite covering source-to-target row counts, null and duplicate checks, referential integrity between patients and encounters, and date and range rules.
Seeded known defects into the dataset to verify the suite detected each failure type, and documented results in a test plan, test cases with expected results, and a defect log.
Ran the suite with Pytest in GitHub Actions CI so any failed data-quality check blocks a merge.