I am a backend engineer with 2 years of professional experience building web APIs, data pipelines, and developer tooling. Most recently at Hoichoi (10M+ users), I worked across our Next.js web application, Go subscription services, and edge deployments on Cloudflare Workers.

In open source, I maintain projects under Nexoral, including AxioDB (an embedded NoSQL database with 20K+ downloads/year) and NexoralDNS (a self-hosted DNS resolver). I enjoy exploring Node.js internals, socket communication, and writing clean, reliable backend code.

• Full Stack Developer · Hoichoi · Jul 2025 – Mar 2026

  • Migrated the Next.js web application from Vercel to Cloudflare Workers using OpenNext, cutting monthly compute costs by $3,000.
  • Implemented retention and cancellation flows in the Go subscription service, integrating Churnkey webhooks and exposing GraphQL queries via NestJS.
  • Fixed notification race conditions by leveraging MongoDB Change Streams to publish to AWS SQS only after database writes commit.

• Software Engineer · Openweb Solutions (Pitangent Group) · Sep 2024 – Jul 2025

  • Handled RTSP camera feeds and WebSocket alert streams in Node.js for an AI video analytics dashboard.
  • Packaged services into Docker containers and configured deployment pipelines to AWS ECS on Fargate.

• Junior Software Developer · Excellis IT · Apr 2024 – Aug 2024

  • Added exponential backoff reconnect logic to prevent MQTT dropouts across 200+ smart lock devices.
  • Set up path-based change detection in GitHub Actions to avoid rebuilding untouched microservices.

Full details → resume.ankan.in

The Problem: SQLite requires native C bindings, creating platform-specific compile issues in Electron apps and requiring node-gyp. Raw JSON storage lacks indexing, transactions, and crash resilience. Client-server databases (like MongoDB) require an external daemon.

Architecture & Implementation:

• Pure TypeScript engine with zero native dependencies; runs directly inside the Node.js/Electron host process.
• Implements MongoDB-compatible query syntax, hash-indexed lookups, and ACID transactions backed by a Write-Ahead Log (WAL).
• Multi-core processing utilizing Node.js worker_threads for parallel dataset operations.
• Ships with an interactive Go CLI, web control dashboard, Dockerized TCP server mode, and a 32-tool Model Context Protocol (MCP) server for direct AI agent interaction.
• Verified with 10 automated test suites covering crash recovery under unexpected SIGKILL, TCP/TLS authentication, and data rollbacks.

Scale & Impact: 20,000+ NPM downloads/year. Used across Electron desktop software and local-first tooling.

The Problem: Managing DNS records and ad/threat filtering across LAN devices typically requires maintaining decentralized /etc/hosts configurations, with zero query audit trails and external DNS providers inspecting outbound queries.

Performance & Architecture:

• The Go Rewrite: Prototyped the initial engine in TypeScript (benchmarked at 8,050 QPS with 0.97% packet loss under 500 concurrent connections). Re-engineered in Go, pushing resolution to 12,746 QPS at 3.8ms latency with zero dropped packets under dnsperf on identical hardware (AMD Ryzen 5 5500U, 9 query worker threads).
• Asynchronous Audit Logging: Shifted query logging and metrics capture to RabbitMQ background consumers after sending the DNS response packet, keeping resolution latency strictly minimal.
• In-Memory Caching: Multi-tier caching layer (in-memory + Redis) absorbs 98% of repetitive lookup traffic.
• Control Interface: Includes a Next.js administrative console and an integrated MCP server for programmatic natural-language DNS rule configuration.

The Problem: Managed cloud load balancers (such as AWS ALB at ~$22/month idle) impose prohibitive fixed base costs for distributed side projects and micro-deployments.

Architecture & Cost Engineering:

• A control plane SaaS that compiles and deploys production load balancers and API gateways directly onto Cloudflare Workers across 330+ edge locations in under 90 seconds.
• Provides 7 routing strategies (weighted, failover, round-robin, etc.), active health checks, JWT validation, and canary releases.
• Cost Efficiency: Reduces idle load-balancing costs to zero by executing within Cloudflare's serverless edge invocation model.
• Optimized AI Tooling: Incorporates a LangChain.js deployment assistant. Dynamically scoped tool definitions decreased token consumption by 66% (from 4,788 down to 1,612 tokens/run), with real-time deployment status streamed via Server-Sent Events (SSE).

• Languages: TypeScript, JavaScript, Go (Working Knowledge), SQL
• Backend & APIs: Node.js, Fastify, Express.js, NestJS, REST APIs, GraphQL, WebSockets
• Databases & Caching: PostgreSQL, MongoDB, Redis, SQLite
• Queues & Asynchronous Workflows: RabbitMQ, AWS SQS, MongoDB Change Streams
• Deployment & Tools: Docker, Linux, Cloudflare Workers, GitHub Actions CI/CD, AWS (ECS, S3)
• Protocols & Standards: DNS (UDP/TCP), MQTT, Model Context Protocol (MCP)