ABUSALEM

With great code comes great responsibility

ABUSALEM

BACKEND SOFTWARE ENGINEER

I was the first engineer to join Twinleaves (now Pallet). Back then, we had no backend — so I rolled up my sleeves, built our first 6 core services from scratch, and set up our cloud foundation. Today, I lead backend engineering across 11 microservices, spending my days tuning search pipelines, building sub-20ms calculation engines, and making sure thousands of stores run without a hitch.

See my work

Chapter 01

ORIGIN STORY

In September 2022, right after finishing intensive backend training at Masai School, I joined Twinleaves (now Pallet) as employee #1 in engineering. There was no existing codebase, no CI/CD pipelines, and no cloud setup — just a big roadmap and a whiteboard. I took total ownership from day one: drafting schemas, writing the API specs, wiring VPC networks on GCP, and pushing our very first commits to production.

During those early startup days, speed and durability went hand in hand. I built 6 of our foundational services from the ground up — Cart, Catalog, Serviceability, Logistics, Communications, and Vendor Management. I also set up our entire GCP infrastructure — Cloud Run, Cloud SQL, VPC networking, Cloud NAT, and automated build pipelines — ensuring we could scale rapidly without outgrowing our architecture.

Fast-forward to today: I lead a talented backend team maintaining 11 microservices. Most of my day is spent solving tricky engineering problems: optimizing an Elasticsearch catalog that handles 45,000+ queries every minute at under 18ms, writing a DAG tax calculation engine that crunch 100K+ daily orders with penny-level precision, keeping checkout latency under 112ms across 1,000+ stores, and shipping practical AI assistants that save merchants hours of manual cataloging.

When I step away from the terminal, I'm usually diving into distributed systems literature, grinding through complex DSA puzzles (700+ solved across LeetCode), or tinkering on side projects like JobBeeper — a tool I built to help job seekers get instant opening alerts across 25+ career boards.

4+

Years Experience

40+

Repositories Contributed

11

Microservices Led

1M+

Catalog Products

Chapter 02

THE ARSENAL

  • Java / Spring Boot96%
  • Distributed Systems & Architecture92%
  • Elasticsearch & Search Pipelines90%
  • PostgreSQL / PostGIS / JPA90%
  • GCP Cloud Infrastructure & CI/CD88%
  • Event-Driven Systems & AI Agents86%

Languages

  • Java
  • SQL
  • JavaScript
  • TypeScript
  • Python

Backend

  • Spring Boot
  • Spring Security
  • REST APIs
  • Microservices
  • JPA / Hibernate
  • Node.js

Data

  • PostgreSQL
  • PostGIS
  • MySQL
  • Elasticsearch
  • Redis
  • BigQuery

Cloud & Messaging

  • GCP (Cloud Run, Cloud Build, IAM, VPC)
  • AWS
  • Pub/Sub
  • Kafka
  • Firebase
  • WebSockets
  • SSE

DevOps & Tools

  • Docker
  • Kubernetes
  • Jenkins
  • GitHub Actions
  • Git

Architecture & Design

  • Multi-tenancy
  • RBAC
  • Event-driven Systems
  • DAG Engines
  • Caching
  • Observability
  • Lock-free Concurrency

Chapter 03

THE WORK

#01

CATALOG & DISCOVERY ENGINE

1M+ products, 45K+ QPM, p99 < 18ms

When your catalog exceeds 1 million products and merchants update stock in real time, basic database lookups break down. I engineered a dedicated catalog engine combining PostgreSQL with custom Elasticsearch query pipelines. It handles multi-tenant discovery, fuzzy search, and dynamic pricing with cache-friendly filter execution. I also designed a multi-level sync engine that automatically fans out updates across hundreds of franchise stores without overriding individual store prices, paired with a backend-driven CMS so the mobile app can change layouts on the fly.

  • Java
  • Spring Boot
  • Elasticsearch
  • PostgreSQL
  • Multi-tenancy
#02

TAX MASTER & DAG ENGINE

100K+ calculations/day, p99 < 20ms

Retail tax in multi-state commerce is full of tricky edge cases — state versus central rules, slabs, cess, cascading percentages, and rounding splits. Instead of hardcoding rules, I designed a zero-code-change Tax Master where finance teams configure rules, jurisdictions, and effective dates. Under the hood, a DAG (Directed Acyclic Graph) engine evaluates dependencies in parallel, computing penny-accurate taxes on 100,000+ daily orders in under 20ms.

  • Java
  • Spring Boot
  • DAG Engine
  • Tax & Billing
  • PostgreSQL
#03

SHARED CART & PROMOTIONS

1,000+ stores, 112ms latency, 5k-10k QPS

A customer adding an item to their cart expects an immediate response, but behind that click are inventory checks, tier pricing, and promo rules. I rebuilt our shared cart engine powering POS, Restaurant, and Consumer apps, slashing add-to-cart latency from 2 seconds down to 112ms under 10,000 queries/sec. I also built a rule-pooling promotion engine that evaluates complex Buy-X-Get-Y offers, bundles, and step discounts on every cart update.

  • Java
  • Spring Boot
  • High-Throughput
  • Promotions
  • Redis
#04

AI AGENTS & COMMONS

70-90% onboarding time reduction + lock-free pool

Onboarding a grocery merchant used to take days of tedious spreadsheet uploads. I built conversational AI catalog and store-builder agents that generate product catalogs, categories, banners, and SEO metadata directly through chat. To keep agent response times fast and economical, I authored an AI commons library featuring a custom lock-free CAS JDBC connection pool that shaved token verification latency down to 5–11ms.

  • Vertex AI
  • Spring Boot
  • SSE
  • Pub/Sub
  • Lock-free JDBC
#05

CLOUD PRINTER & JOB SYSTEM

50K–80K print jobs/day across 400+ stores

Shop floors and commercial kitchens have notoriously spotty Wi-Fi, yet billing receipts and Kitchen Order Tickets (KOT) can never be lost. I built a centralized job orchestration engine and lightweight Cloud Printer Agent that reliably handles 50K–80K daily print jobs. It uses duplex WebSockets with automatic HTTP polling fallback, startup spool recovery, and reprint tokens so no order is ever missed.

  • WebSockets
  • Kafka
  • Pub/Sub
  • POS
  • Event-driven
#06

JOBBEEPER

Personal project · 25+ platforms, 2K+ active users

Job searching is stressful when you have to refresh 20 different portals every day. I built JobBeeper to aggregate openings across 25+ major job boards (LinkedIn, Naukri, Workday, Greenhouse, Lever, Ashby) into a unified stream. It runs intelligent matching, deduplication, and retries, sending instant personalized alerts via Telegram, Web Push, and Android. It runs 24/7 on Oracle Cloud with Docker, PostgreSQL, Redis, and Celery workers.

  • Python
  • PostgreSQL
  • Redis
  • Docker
  • Nginx
Live→

Chapter 04

THE HISTORY

  1. Twinleaves (Pallet)

    first engineering employee · grew platform from 0 to 11 microservices

    Sep 2022 — Present · Bengaluru, India

    1. Software Development Engineer 2

      May 2024 — Present
      • Catalog & Discovery: Engineered high-throughput catalog search and hierarchy engine (PostgreSQL + Elasticsearch) for 1M+ products, handling 45K+ queries/min with p99 search latency under 18ms.
      • Elasticsearch Pipelines: Built custom Elasticsearch query pipelines for multi-tenant product discovery, dynamic filtering, aggregations, and pricing with cache-friendly filter execution and query optimization.
      • Catalog Synchronization: Built a multi-level catalog sync engine that propagates master product changes across hundreds of stores while preserving store-specific pricing and inventory overrides.
      • Backend-Driven Content: Designed and built a backend-driven app content system to manage pages, layouts, and campaigns with scheduled festival and promotion updates without requiring new client releases.
      • Tax Master & DAG Engine: Built a configurable Tax Master (jurisdictions, rules, rates, slabs, exemptions) and a high-precision DAG-based tax calculation engine processing 100K+ calculations/day with p99 latency under 20ms and rounding reconciliation.
      • Cart & Promotions: Built a shared cart service across 1,000+ stores for POS, Restaurant, B2C, and B2B flows. Reduced add-to-cart latency from 2s to 112ms under 5k–10k+ QPS, and built a promotion engine for Buy-X-Get-Y, step offers, combos, and category pooling.
      • AI Agents: Built AI catalog and website builder agents that manage products, SEO, layouts, and styling through chat with inline actions, slashing store setup and onboarding times by 70–90%.
      • Platform & AI Commons: Built a shared Spring Boot platform library for multi-tenant auth, RBAC, and rate limiting; built an AI commons library with a lock-free CAS JDBC connection pool achieving sub-millisecond borrow latency and reducing AI token checks to 5–11ms.
      • Job Management & Cloud Printing: Built a centralized Job System and Cloud Printer Agent processing 50K–80K print jobs/day across 400+ stores using WebSockets, polling fallback, and Pub/Sub + Kafka event streams.
      • Serviceability & Logistics: Built a configurable serviceability service across 1,000+ stores for instant, same-day, and slotted delivery using PostGIS for geo-radius and delivery-time calculation under 45ms.
      • Cloud & Infrastructure: Managed production GCP infrastructure across 110+ Cloud Run services, 10+ Compute Engine instances, 10+ Cloud SQL instances, 170+ Artifact Registry repos, 140+ Cloud Build triggers, 70+ Pub/Sub topics, and end-to-end VPC networking.
      • Technical Leadership: Lead 3 backend engineers across 11 microservices (specs, PR reviews, releases); previously led a 14-member cross-functional team (backend, frontend, QA) for Pallet Restaurant across RMS, POS, B2C, B2B, and Zomato/Swiggy integrations.

    2. Software Development Engineer 1

      Sep 2022 — May 2024
      • Core Services: Joined as the first engineering hire and independently built 6 core services from scratch: Cart, Catalog, Serviceability, Logistics, Communications, and Vendor Management, while contributing to 10+ others.
      • Integrations & Offline POS: Integrated WhatsApp Business, GS1 Datakart, Firebase, and Vertex AI; engineered offline POS inventory synchronization to maintain uninterrupted store operations during network outages.
      • GCP Foundation: Set up the company's entire GCP cloud infrastructure from the ground up, including Cloud Run, Cloud SQL, VPC networks, subnets, firewalls, Cloud NAT, load balancers, Docker, and Cloud Build CI/CD.

  2. Training

    • Java Backend Development

      Masai School, Bengaluru

      Dec 2021 — Aug 2022

    • Diploma in Computer Science

      Bharati Vidyapeeth, Maharashtra

      Aug 2018 — Oct 2021

Chapter 05

THE FEATS

  • 40+

    Repositories Contributed

    Across core backends, microservices, CLI agents & UI

  • 112ms

    Add-to-cart latency

    Reduced from 2s under 10k QPS

  • < 18ms

    p99 search latency

    1M+ products, 45K+ QPM

  • 100K+

    Daily tax calculations

    DAG engine with p99 < 20ms

  • 1,000+

    Stores powered

    POS, Restaurant, B2C & B2B

  • 50K–80K

    Daily print jobs

    Cloud Printer Agent across 400+ stores

  • Employee of the Quarter

    Awarded every quarter since December 2022 at Twinleaves (Pallet). Joined as the first engineering employee and now lead backend development across 11 microservices.

  • 700+ DSA Problems Solved

    Solved 700+ data structures and algorithms problems across LeetCode and other competitive programming platforms.

Platform Footprint40+ Repositories

Systems & Repositories Contributed

From zero-to-one microservices to real-time shop-floor agents, shared platform starters, and automated test suites — here is a snapshot across the repository ecosystem I helped build and scale.

Core Commerce & High Scale

  • cmsCatalog & App Content Engine

    Elasticsearch indexing, multi-level hierarchy, 45K+ QPM

  • pallet-cartShared Cart Engine

    Multi-tenant cart, promotions, 112ms p99 latency

  • pallet-tax-engineTax Master & DAG Engine

    100K+ daily calculations, sub-20ms DAG pipeline

  • pallet-serviceabilityGeo-Serviceability Engine

    PostGIS instant/slotted delivery calculation < 45ms

  • plusmax-cmsMerchant App Content

    Dynamic layouts, banner feeds, festival updates

  • plusmax-companyCompany & Multi-tenant Config

    Store hierarchies, merchant profiles, tenant policies

Store Infrastructure & Real-Time Ops

  • pallet-printerCloud Printer Agent

    WebSocket + polling fallback, 50K-80K print jobs/day

  • pallet-pos-servicePOS Cloud Gateway

    Billing, offline cash registers, token syncing

  • pallet-comms-agentCommunications Agent

    Background SMS, WhatsApp Business, transactional mailers

  • pallet-highwayLogistics Highway

    Rider dispatch, slotted order fulfillment pipelines

  • vmsVendor Management System

    Purchase orders, supplier catalog matching, ledger syncing

  • offline-pos-agentOffline POS Agent

    Local SQLite caching & auto-sync during internet dropouts

AI, Tools & Platform Starters

  • pallet-ai-commonsAI Commons Library

    Lock-free CAS JDBC pool, token metering (5-11ms)

  • pallet-mcp-serviceModel Context Protocol Service

    AI assistant tools for live store data & commands

  • pallet-chatbotAI Merchant Assistant

    Natural language catalog & promotion builder agent

  • twinleaves-common-starterCompany-wide Spring Starter

    Standardized multi-tenant auth, filters, error handling

  • pallet-commons-codeShared Utilities & DTOs

    Core data contracts shared across 11 microservices

  • pallet-cronCentralized Scheduler

    Automated reconciliation, inventory audits, billing triggers

Frontend, Portals & Automated Testing

  • pallet-rms-uiRestaurant Management Portal

    Table management, KDS (Kitchen Display), menus

  • pallet-mrms-replitMerchant Mobile RMS

    Mobile-first merchant dashboard & real-time analytics

  • pallet-websitePallet Public Platform

    High-performance marketing & product portal

  • pallet-grocery-web-uiB2C Consumer Web Store

    E-commerce storefront for supermarket chains

  • web-pos-playwrite-automationE2E POS Test Suite

    Automated regression testing for POS checkout journeys

  • plusmax-erp-automationERP Integration Suite

    Automated reconciliation against external ERP ledgers

Chapter 06

THE TIMELINE

A week-by-week executive summary of systems engineered, architectures scaled, and platform milestones delivered at Pallet / Twinleaves.

07 Sep 2026Week 37 · 2026

Cart Promotion Engine & Bundle Discount Matching

Cart & Promotions
  • Handled same-category and same-brand Buy X Get Y offers across individual products and pooled leftovers
  • Rejected offers when the threshold was nonpositive or the cart quantity was below it
31 Aug 2026 – 02 Sep 2026Week 36 · 2026

Cart Promotion Engine & Bundle Discount Matching

Catalog & SearchCart & Promotions
  • Calculated flat-price discounts using the number of eligible bundles instead of multiplying the flat price by quantity
  • Changed the Docker builder to assemble the application without running verification tasks
25 Aug 2026 – 28 Aug 2026Week 35 · 2026

Catalog Hierarchy Sync & Multi-tenant Store Overrides

Catalog & Search
  • Aligned governed write and approval flows with consistent replay, transaction, and response handling
  • Updated product filtering to match governance values stored as either direct fields or keyword subfields
17 Aug 2026 – 19 Aug 2026Week 34 · 2026

Elasticsearch Indexing & Query Pipeline Optimization

Catalog & Search
  • Corrected legacy date-window upper bounds and made the Elasticsearch sequence seed use the largest available legacy numeric identifier
  • Renamed routing and approval types to reflect their actual responsibilities and synchronized PME structure documentation
12 Aug 2026 – 13 Aug 2026Week 33 · 2026

Elasticsearch Indexing & Query Pipeline Optimization

Catalog & Search
  • Design spec for grocery bulk upload variants and restaurant variant fixes
  • Make apply/merges the whole merge pipeline and self-approve what qualified
03 Aug 2026 – 06 Aug 2026Week 32 · 2026

Catalog Hierarchy Sync & Multi-tenant Store Overrides

Catalog & Search
  • Pme: the three backfill controllers become one, and take a landmine with them
  • Pme: cross-level product filter — the v2 contract answered at every level
27 Jul 2026 – 31 Jul 2026Week 31 · 2026

AI Catalog Agent & Automated Merchant Onboarding

Catalog & SearchCart & PromotionsMerchant Web UI
  • Pme G3: the whole rollout rehearsed on a copy of RET_82, and the readiness count that could never reach zero
  • Pme C3: the name cascade wrote every vertical into grocery, and stamped ids on sites that never matched
20 Jul 2026 – 21 Jul 2026Week 30 · 2026

AI Catalog Agent & Automated Merchant Onboarding

Catalog & Search
  • Default the HO tier to the org node when no group id is known
  • Make chains account-optional and preserve explicit ho/brand ids

Chapter 07

SEND UP THE SIGNAL

Whether you're tackling high-throughput search, architecting low-latency transactional engines, or looking for a hands-on technical lead to take systems from zero to one — I'm always up for a good engineering conversation. Drop me an email or find me on the channels below.

mail@abusalem.me

Based in Bengaluru, India