Atul SinghSenior Technical Lead · Appinventiv
I design systems that scale.
Backend engineer & system designer — I turn ambiguous requirements into event-driven systems that stay calm under real load.
7+Years ExperienceBackend & architecture
1M+Events / 5 minReal-time pipelines →
Noida, INOpen to Remote
Focus Areas
Selected impact
- 300xAPI latency cut
- 94.7%cache hit rate
- 10+engineers led
- 150+tests on a split monolith
01. What I Work With
Architecture & Design
- Distributed SystemsCoreDesigning for partial failure, consistency, and scale across services.
- System DesignCoreEnd-to-end design of large-scale, high-throughput platforms.
- High-Level DesignCoreService boundaries, data flow, and component decomposition.
- Low-Level DesignCoreClass/module design, interfaces, and clean design patterns.
- CQRSCoreCommand Query Responsibility Segregation for write/read separation.
- Event-Driven ArchitectureCoreLoosely coupled event channels and message flow orchestration.
- MicroservicesProductionDecoupled domain services designed around boundaries.
- Domain-Driven DesignCoreDesigning architecture around clean business subdomains.
- API DesignCoreType-safe interfaces designed for longevity and clarity.
- gRPCProductionHigh-performance RPC using Protocol Buffers.
AI & LLM Engineering
- Large Language Models (LLM)ProductionBuilding production features on top of LLM providers.
- RAGProductionRetrieval-Augmented Generation over vector-indexed knowledge.
- Model Context Protocol (MCP)CoreStandardized tool/context interface for AI agents.
- LangChainProductionOrchestration framework for LLM chains and agents.
- pgvectorProductionVector similarity search inside PostgreSQL for embeddings.
- Semantic CachingProductionEmbedding-based cache to cut redundant LLM calls.
Backend & Infrastructure
- Node.jsCoreCore daily driver runtime for event-driven backend APIs.
- TypeScriptCoreType-safety layer ensuring scalable codebase consistency.
- PythonProductionData analytics tooling and script orchestration.
- NestJSCoreStructured framework matching MVC/DDD design patterns.
- RESTful APIsProduction
- GraphQLProductionFlexible querying API layer for clientside offloading.
- FastAPIProduction
Data Architecture
- PostgreSQL (MVs)ProductionMaterialized Views scaling, handling 1M+ events / 5min at Appinventiv.
- MongoDBProductionDocument storage used for flexible document catalog schemas.
- RedisCoreHigh-performance key-value distributed caching.
- ElasticsearchProductionInverted index full-text search engine scaling.
- GIN IndexesCoreGeneralized Inverted Index for fast array queries.
Cloud & DevOps
- Azure (Event Hub, Blob Storage)ProductionMain cloud pipeline context for real-time ingestion.
- AWS (S3, SQS, Lambda, EC2)ProductionStandard serverless, messaging, and cloud deployments.
- KubernetesProductionContainer orchestration across multi-zone nodes.
- DockerCoreIsolated local containerization of domain microservices.
- Apache AirflowProduction
Security & Auth
- OAuth2.0CoreStandard auth protocol for federated credentials.
- JWTCoreSecure payload transfer using cryptographically signed tokens.
- RBACProduction
- Entra IDProduction
- KeycloakProduction
CI/CD & Tools
- GitHub ActionsProduction
- GitLab CIProduction
- JIRAProduction
- Agile/ScrumProduction
Core Expertise
- Distributed SystemsDesigning for partial failure, consistency, and scale across services.
- System DesignEnd-to-end design of large-scale, high-throughput platforms.
- High-Level DesignService boundaries, data flow, and component decomposition.
- Low-Level DesignClass/module design, interfaces, and clean design patterns.
- CQRSCommand Query Responsibility Segregation for write/read separation.
- Event-Driven ArchitectureLoosely coupled event channels and message flow orchestration.
- Domain-Driven DesignDesigning architecture around clean business subdomains.
- API DesignType-safe interfaces designed for longevity and clarity.
- Model Context Protocol (MCP)Standardized tool/context interface for AI agents.
- Node.jsCore daily driver runtime for event-driven backend APIs.
- TypeScriptType-safety layer ensuring scalable codebase consistency.
- NestJSStructured framework matching MVC/DDD design patterns.
- RedisHigh-performance key-value distributed caching.
- GIN IndexesGeneralized Inverted Index for fast array queries.
- DockerIsolated local containerization of domain microservices.
- OAuth2.0Standard auth protocol for federated credentials.
- JWTSecure payload transfer using cryptographically signed tokens.
Production Stack
- MicroservicesDecoupled domain services designed around boundaries.
- gRPCHigh-performance RPC using Protocol Buffers.
- Large Language Models (LLM)Building production features on top of LLM providers.
- RAGRetrieval-Augmented Generation over vector-indexed knowledge.
- LangChainOrchestration framework for LLM chains and agents.
- pgvectorVector similarity search inside PostgreSQL for embeddings.
- Semantic CachingEmbedding-based cache to cut redundant LLM calls.
- PythonData analytics tooling and script orchestration.
- RESTful APIs
- GraphQLFlexible querying API layer for clientside offloading.
- FastAPI
- PostgreSQL (MVs)Materialized Views scaling, handling 1M+ events / 5min at Appinventiv.
- MongoDBDocument storage used for flexible document catalog schemas.
- ElasticsearchInverted index full-text search engine scaling.
- Azure (Event Hub, Blob Storage)Main cloud pipeline context for real-time ingestion.
- AWS (S3, SQS, Lambda, EC2)Standard serverless, messaging, and cloud deployments.
- KubernetesContainer orchestration across multi-zone nodes.
- Apache Airflow
- RBAC
- Entra ID
- Keycloak
- GitHub Actions
- GitLab CI
- JIRA
- Agile/Scrum
Actively Exploring
- Agentic WorkflowsMulti-step autonomous agents with tool use and planning.
- OpenTelemetryDistributed tracing and unified observability pipelines.
- Temporal.ioDurable workflow orchestration for long-running processes.
02. Where I've Worked
Senior Technical Lead @ Appinventiv
Jul 2019 - Present · 7 yrs
Career Progression at Appinventiv:
Senior Technical Lead (Aug 2026 – Present)
Technical Lead (2024 – Aug 2026)
Associate Tech Lead (Mar 2022 – 2024)
Senior Software Engineer (Mar 2020 – Mar 2022)
Software Engineer (Jul 2019 – Mar 2020)
Project: Quivio (CRM & Core Platform)
Team Leadership & Scale:
- Led a team of 10+ backend engineers using Agile/Scrum methodologies, conducting Sprint Planning, Code Reviews, and System Design sessions via JIRA
- Spearheaded architecture of a multi-tenant SaaS platform processing 1M+ events per 5 minutes, enabling real-time marketing automation at scale
Performance Optimization (300x-600x improvements):
- Transformed customer segmentation API from 30s to 100ms (300x faster) by designing PostgreSQL Materialized Views with GIN-indexed arrays and statement-level triggers for real-time notification
- Implemented a Migration-Aware Refresh strategy to handle daily bulk data imports (1M+ rows), utilizing debouncing and window-based suppression to maintain system stability during heavy loads
- Slashed query latency from 3000ms to 30ms (100x improvement) by eliminating N+1 patterns through batch processing with Set-based lookups
- Drove 65% memory reduction and 94.7% cache hit rate by implementing LRU-based tenant connection pooling with reference counting
Event-Driven Architecture & Security:
- Engineered high-throughput Event-Driven Architecture pipeline using Azure Event Hub (32 partitions) with BlobCheckpointStore, achieving 10x capacity headroom and zero-downtime partition rebalancing
- Implemented production-grade resilience: exponential backoff (1s/5s/15s), Redis-based duplicate detection, auto-reconnect health checks, and Kubernetes HPA autoscaling (10-20 pods)
- Implemented OAuth2.0 authentication and RBAC access control using Entra ID (Azure AD) and Keycloak, with JWT token management
Code Quality & Refactoring:
- Refactored 6,634-line monolithic service into 6 focused microservices, increasing test coverage to 150+ unit tests and reducing deployment risk
Project: ECCO (E-Commerce Platform)
- Led a team of 2 backend developers delivering scalable RESTful APIs and GraphQL e-commerce APIs
- Built GraphQL APIs on Commercetools for product catalog and cart management
- Designed Backend-for-Frontend (BFF) architecture serving web, Android, and iOS clients with optimized data payloads
- Integrated Algolia for sub-100ms search and Contentful CMS for dynamic content
Senior Software Engineer Initiatives:
- Architected real-time Event-Driven Architecture communication system using Redis Pub/Sub
- Designed streaming pipelines from MongoDB to Firebase push notifications following SOLID Principles
- Built seat management system for co-working spaces with optimized allocation workflows and RESTful APIs
- Implemented CI/CD pipelines using GitHub Actions for automated testing and deployment
Tech Stack: Node.js, TypeScript, NestJS, PostgreSQL, MongoDB, Redis, Azure Event Hub, Azure Blob Storage, AWS (S3, SQS), Kubernetes, Apache Airflow, GraphQL, GitHub Actions, JIRA
03. Things I've Built
Multi-tenant SaaS CRM platform for car wash operations, processing 1M+ events per 5 minutes across 32 Azure Event Hub partitions. Architected the event pipeline with BlobCheckpointStore and Kubernetes HPA (10–20 pod autoscaling) to achieve 32K events/sec burst capacity — 10x headroom over peak load. Reduced customer segmentation API latency from 30s to 100ms (300x) using PostgreSQL Materialized Views with GIN-indexed arrays. Led a team of 10+ backend engineers through decomposition of a 6,634-line monolith into 6 focused microservices with 150+ tests.
Amex Sadad Payment Gateway
AppinventivDesigned a vendor-agnostic Payment SDK for American Express to integrate Sadad’s payment gateway without operational coupling to Sadad’s release cycle. Architected the service-to-service contract using gRPC with Protocol Buffers — a stable versioned interface abstracting Sadad’s internal data model behind a boundary Amex could own. Chose Oracle DB to satisfy Amex’s audit infrastructure requirements. Coordinated technical specifications across two enterprise organisations with distinct governance and compliance mandates.
E-commerce platform for ECCO footwear built on Commercetools, serving web, Android, and iOS through a Backend-for-Frontend (BFF) architecture that delivers optimised payloads per client type. Integrated Algolia (sub-100ms search) and Contentful CMS, decoupling marketing content updates from the engineering release cycle. Led 2 backend developers building GraphQL APIs for product catalog, cart, and checkout flows.
Financial super app by Unitel (Mongolia’s leading telecom) aggregating parking, taxi, mobile top-ups, and utility payments. Built on a microservices topology with PostgreSQL handling transactional data and MongoDB handling user profiles and activity feeds — chosen by write-pattern, not convention. Designed inter-service communication to handle partial failures explicitly: a taxi booking failure must not silently roll back an unrelated top-up transaction.
Cloud-based SaaS for corporate workspace management enabling seat reservation, occupancy analytics, and space optimisation. Designed the booking data model with conflict-free concurrent reservations using optimistic locking, eliminating double-booking edge cases at peak check-in hours. Built real-time occupancy tracking via event-driven state propagation.
Food ordering platform serving Paradise Biryani outlets nationwide. Designed the caching layer using Redis and AWS ElastiCache for menu and inventory reads, reducing MongoDB load during peak ordering hours. Built multi-outlet data isolation while maintaining a unified customer ordering experience.
AI-Powered Business Intelligence Bot that transforms natural language into SQL queries. Built with a 3-Layer Search Engine architecture (Semantic Cache → RAG → LLM), reducing recurring text-to-SQL query latency from 1.5s to <20ms. Features intelligent visualization that dynamically renders SQL results into KPI cards or charts based on data shape analysis.
A production-ready boilerplate for building Node.js applications with TypeScript and Express. Includes best practices for project structure, error handling, validation, and testing setup.
A high-performance, scalable REST API for e-commerce applications. Designed with microservices principles, featuring advanced caching, database optimization, and containerization for easy deployment.
Implementation and scaling strategies for Node.js job queues using the Bull library. Focuses on handling high-throughput background jobs, reliability, and horizontal scaling with Redis.
04. How I Think
As a Solution Architect, my role is to make structured trade-offs. I document these choices using Architecture Decision Records (ADRs) based on four core tenets:
01
Bounded Contexts
Domain boundaries dictate service splits, not technology layers. Keep microservices aligned with business subdomains to reduce contract drift.
02
Bounded Latency
OLTP endpoints must respond under 100ms. Offload analytics, reporting, and complex segmentation queries to pre-computed materialized views.
03
Event-Driven Decoupling
Use messaging brokers (Azure Event Hub, RabbitMQ) to buffer traffic spikes. Guarantee message processing with dead-letter loops and idempotent receivers.
04
Observability First
If a metric isn't tracked, it doesn't exist. Implement structured correlation IDs across all logs, traces, and metrics before launching to production.
ADR #04 CASE STUDY
Materialized Views vs. Distributed Cache
Scenario: Live segment builder count counts millions of rows with AND/OR customer filters.
Pre-Computed Architecture Flow
flowchart TD A[Customer Writes] -->|Transactional DB| B[(PostgreSQL DB)] B -->|Statement Trigger| C[MV Refresh Queue] C -->|Atomic REFRESH CONCURRENTLY| D[(Materialized View)] E[API Filter Request] -->|Index-only Scan| D D -->|Sub-100ms Response| E
| Attribute | Option A: PostgreSQL MVs (Selected) | Option B: Redis Cache Grid |
|---|---|---|
| Read Latency | 50ms - 200ms (Covering GIN Index) | 2ms - 10ms (In-Memory Key) |
| Data Invalidation | Automatic triggers on base table write | Complex TTL eviction & sync handlers |
| Storage Overhead | Low (Fits in standard index memory heap) | Very High (Holds complete datasets in RAM) |
| Query Flexibility | Infinite combinations (AND, OR, arrays) | Extremely Limited (Key-value structure only) |
Decision: Selected Materialized Views. While Redis offers faster raw read speeds, the complex multi-select filter combinations (65+ attributes) would require storing millions of combinatorial key sets in memory. PostgreSQL's GIN indexing provides clean array-matching query flexibilities directly inside the transactional database heap.
05. Featured Writing
06. What People Say
Atul was such a great engineer to work with. We have worked together on numerous complex systems and his problem solving skills and determination were a key effort to the overall success of the project we worked on.
I was Atul's client for about a year. He is an experienced Tech Lead with all the soft and hard skills needed to succeed in his position. Very reliable and easy going!
Working with Atul was a pleasure. He is a very senior developer and people leader — really consistent, focused, and always giving the extra mile.
07. What's Next?
Get In Touch
I’m open to senior engineering leadership and solution architecture opportunities. If you’re building a complex distributed system and need someone who can own the architecture, lead the team, and ship it — I’d love to talk.