Skip to content

Case study index · 07

Backend & data engineering projects

Project breakdowns with architecture choices, measurable targets, and engineering tradeoffs.

  1. 01live forward test

    NEPSE Market Prediction System

    End-to-end, self-hosted ML system for the Nepal Stock Exchange: 30 years of cleaned market data, a walk-forward XGBoost model scoring every stock after each session, served through FastAPI and a React dashboard. Built to be audited, not hyped.

    Canonical rows
    511,274
    Rank IC
    0.093
    Tests
    ~720
    • Python
    • XGBoost
    • pandas
    • FastAPI
    • React
    • Playwright
    • Docker
    • Parquet
    nepseportalsmirroringestparquetfeaturesxgboostfastapiaudit log
  2. 02live· Techspire College

    Techspire College Website

    End-to-end build of Techspire College's public website, from system architecture and information architecture through wireframes to a production launch.

    Scope
    Architecture → launch
    Role
    Architect & lead dev
    Audience
    Students & parents
    • Astro
    • React
    • System Architecture
    • Information Architecture
    • Wireframing
    architectureinfo archwireframesui buildlaunchadmissions
  3. 03in production· Techspire College

    Student Onboarding Automation

    A GitHub-driven student onboarding form for Techspire College where every submission from the website flows automatically into Google Sheets for the admissions team.

    Flow
    Form → Google Sheets
    Manual entry
    Removed
    Users
    Admissions team
    • GitHub
    • Google Sheets
    • Web Forms
    • Automation
    studentformgithubsyncg-sheetsadmissions
  4. 04completed

    Microservices API Gateway

    Designed for 25,000 req/sec with routing, rate limiting, and observability to contain failure spread.

    Traffic Target
    25,000 req/sec
    Latency Gain
    60% lower
    Reliability
    99.9% uptime
    Daily Data
    500GB
    • Go
    • Redis
    • Prometheus
    • Grafana
    • Docker
    • Kubernetes
    clientgatewayauthordersuserstelemetry
  5. 05completed

    Distributed Twitter Analytics Platform

    Processes 2.5TB/day with streaming workers and 85% lower query latency.

    Traffic Target
    10,000 req/sec
    Latency Gain
    85% lower
    Reliability
    99.9% uptime
    Daily Data
    2.5TB
    • Go
    • gRPC
    • Redis
    • Apache Kafka
    • ClickHouse
    • Docker
    streamkafkaworkerworkerworkerwarehouse
  6. 06completed

    Cloud-Native Auth System

    Supports tenant-aware auth at 8,000 req/sec with RBAC, Redis sessions, and audit controls.

    Traffic Target
    8,000 req/sec
    Latency Gain
    45% lower
    Reliability
    99.95% uptime
    Daily Data
    100GB
    • Node.js
    • TypeScript
    • PostgreSQL
    • Redis
    • JWT
    • Docker
    clientauthrbacredisaudit
  7. 07completed

    Real-Time Data Pipeline Engine

    Unifies streaming and batch ETL for 800GB/day with quality and lineage checks.

    Traffic Target
    5,000 req/sec
    Latency Gain
    70% lower
    Reliability
    99.5% uptime
    Daily Data
    800GB
    • Python
    • Apache Airflow
    • Apache Spark
    • PostgreSQL
    • Docker
    • AWS S3
    apidbeventsairflowsparkqualitys3 lake