Lead backend engineer at dmclounge.ai

Ibuildbackendsthatholdupinproduction.

I'm Bishal, a software engineer with 4+ years shipping and running production systems. Lead backend engineer on a multi-tenant travel SaaS — system design, tenant isolation, real-time Gmail & Outlook, metered LLM features, deployments and the monitoring that keeps it all up.

See my workBook a call
Kathmandu, Nepal· 
platform · event streamlive
$
Illustrative · multi-tenant SaaS · Spring Boot · AWS · GCP
0+
years running production systems
0+
agencies on real-time Gmail & Outlook
~0%
faster scraping pipeline
~0%
faster article search

01 / Experience

Problems I’ve owned, and how I solved them.

From scraping pipelines to the backend of a multi-tenant SaaS — each role framed as the problem, the fix, and what changed.

  1. Jul 2025 — Present

    Remote

    dmclounge.ai

    Software Engineer · Lead backend — Multi-tenant SaaS platform for travel agencies

    • Spring Boot
    • PostgreSQL
    • AWS SQS
    • GCP Pub/Sub
    • Cloud Run
    • Microsoft Graph
    • Gemini
    • WebSocket
    • New Relic

    • Real-time email

      5+ agencies

      Replaced polling with push — Gmail watch via GCP Pub/Sub and Microsoft Graph change notifications for Outlook, per-mailbox OAuth and WebSocket delivery to the UI. New mail lands within seconds.

    • AI reply drafting (RAG)

      Indexes email history into a per-tenant knowledge base and drafts replies from the customer profile and thread context, citing sources so agents can verify before sending.

    • AI cost control

      Built an AI gateway that meters tokens per tenant and per feature, with pay-as-you-go billing across Gemini and Google Speech-to-Text — turning AI usage into a billable feature.

    • Tenant isolation

      Database-per-tenant with request-scoped datasource routing in Spring Boot. Runs on AWS with SQS for email processing and GCP Cloud Run for Gmail ingestion.

    • Memory & CPU issues

      Added per-request CPU-time and memory monitoring reported to New Relic, pinpointing the exact endpoints and jobs behind repeated OOMs and CPU spikes.

    • Observability

      New Relic APM with environment-separated NRQL dashboards and error alerts routed to Slack, so triage starts with the answer instead of the question.

  2. Jul 2024 — Jul 2025

    SecureX Lab

    Software Engineer

    • Spring Boot
    • PostgreSQL
    • Liquibase
    • Next.js
    • TypeScript
    • Xero
    • Django
    • React

    • Inventory & FIFO costing platform

      Turned an import business’s loosely defined needs into a technical proposal, phased plan and working product. Designed the backend on Spring Boot, PostgreSQL and Liquibase.

    • Correct costing

      FIFO stock dispatch and value-proportional landed-cost allocation across a per-product packaging hierarchy. Chose FIFO-only over a mixed model to keep costs traceable and auditable.

    • Reliable accounting sync

      Hourly Xero sync on an outbox / sync-queue pattern — a crash mid-sync is retried safely, never lost or double-posted. Integration-tested against real PostgreSQL.

    • HR & payroll

      Attendance system pulling punch data from biometric devices into salary calculation, with HR dashboards. Used daily by the client’s employees and HR team.

    • Client websites

      Custom business websites to customer requirements, including real-time chat over WebSockets (Django, React).

  3. Jan 2022 — Jun 2024

    Quickfox Consulting

    Associate Software Engineer

    Joined as Software Engineering Intern

    • Python
    • Scrapy
    • Selenium
    • Elasticsearch
    • Transformers

    • Scraping reliability

      ~83% faster

      Proposed and led the move from RPA-driven Chrome instances to Scrapy with headless fetching, running many spiders in parallel. The server stopped locking up.

    • Search

      ~60% faster

      Moved article search from SQL queries to Elasticsearch.

    • Ownership

      Owned the scraping server and daily scrape monitoring; shipped features like a voting system and turned user feedback into UI and API improvements.

    • NLP

      Added transformer-based sentiment classification to categorise news articles automatically.

  4. Education

    BE, Software Engineering

    Nepal College of Information Technology — Pokhara University · Balkumari, Lalitpur

02 / Skills

The stack, in orbit.

Languages at the core, then backend, data, cloud — and the AI, integrations and frontend tooling around them. Hover a group to focus it.