Building scalable backend systems
Hi, I'm Sujal.
Backend Engineer with experience designing scalable systems and optimizing performance. Currently building backend systems at Velocity Clinical Research using Node.js, FastAPI, and AWS.
About Me
Backend Engineer based in Hyderabad with ~2 years of experience building scalable production systems.
Focused on designing reliable backend services, optimizing performance, and engineering clean, resilient APIs.
Experienced in architecting event-driven systems and resolving high-severity production issues.
~2
Years Experience
50+
Production Incidents Resolved
Featured Projects
MedInsight - AI Medical Assistant
AI-powered system for processing medical reports using asynchronous pipelines and LLM-based workflows.
Engineering Highlights
- ●Engineered an asynchronous PDF ingestion pipeline isolating long-running processing from user requests.
- ●Implemented Redis-based rate limiting to ensure system stability under concurrent usage.
- ●Designed a state-machine job processor handling PENDING/FAILED/COMPLETED states with retry mechanisms.
Node.jsTypeScriptReactLangChainZodReact QueryOpenAI APIRedisMongoDB
Architecture Overview
- ●Asynchronous processing pipeline decoupled from user interaction layer.
- ●LLM-based extraction workflow integrated with structured validation using Zod.
Key Engineering Decisions
- ●Decoupled ingestion from request lifecycle to avoid blocking and improve responsiveness.
- ●Used schema validation (Zod) to ensure reliable structured outputs from LLM responses.
Challenges & Solutions
- ●Handling memory-intensive multi-page PDF parsing during OCR and extraction stages.
SyncScript - Collaborative Editor
Real-time collaborative text editor built on WebSockets supporting low-latency synchronization for 50+ concurrent users.
Engineering Highlights
- ●Engineered conflict-resolution logic preventing race conditions across concurrent users.
- ●Optimized WebSocket broadcasting for low-latency real-time synchronization.
- ●Implemented PostgreSQL batching to reduce database load under concurrent edits.
Node.jsWebSocketsReactTypeScriptPostgreSQL
Architecture Overview
- ●WebSocket-based real-time communication layer over Node.js.
- ●Delta-based updates instead of full document sync to improve efficiency.
Key Engineering Decisions
- ●Used operational transformation strategies instead of locking mechanisms.
- ●Batched database writes instead of per-keystroke persistence.
Challenges & Solutions
- ●Handling out-of-order updates caused by network latency and jitter.
Technical Skills
Languages
TypeScript
JavaScript
Python
Java
SQL
Backend
Node.js
Express
FastAPI
WebSockets
Databases & Caching
PostgreSQL
MongoDB
MS SQL Server
Redis
Cloud & DevOps
AWS
Docker
CI/CD
Git
Azure DevOps
AWS Batch
SendGrid
Twilio
Lambda
S3
CloudWatch
System Design
Microservices
Event-driven Architecture
Caching
Rate Limiting
Load Balancing
Experience
Where I've worked and what I've accomplished.
Velocity Clinical Research
Software Engineer
Aug 2024 - PresentKey Projects
Enterprise Analytics Platform
- ●Designed an event-driven microservices architecture to surface data quality insights at scale.
- ●Engineered parallel ingestion pipelines using Node.js and AWS services to move data from EDW to PostgreSQL.
- ●Optimized database concurrency by isolating shared tables, preventing row-locking bottlenecks.
CTA Extraction (LLM Pipeline)
- ●Architected a multi-step LLM pipeline to extract structured data from complex clinical trial PDFs.
- ●Improved extraction accuracy by decomposing processing into page classification and table detection stages.
- ●Leveraged AWS Batch and Lambda to orchestrate asynchronous PDF extraction workflows.
AI Voice Assistant (Twilio)
- ●Built an AI-powered call automation system using OpenAI and Twilio to streamline patient scheduling.
- ●Implemented asynchronous processing workflows to handle high-concurrency call volumes.
System Design Details
Architecture Overview
- ●Microservices-based system with event-driven orchestration.
- ●Parallel ingestion pipelines decoupled from aggregation workflows.
- ●Multi-step asynchronous LLM extraction pipeline.
Key Engineering Decisions
- ●Separated ingestion and aggregation layers to improve scalability.
- ●Used table isolation to eliminate database contention issues.
- ●Decomposed LLM tasks into smaller steps for higher accuracy.
Challenges & Solutions
- ●Handling high-throughput data ingestion alongside complex PDF extraction workflows.
Software Engineering Intern
Aug 2023 - Jul 2024Key Projects
Patient Registration Application
- ●Built a React-based patient registration application with full frontend implementation.
- ●Implemented component-level testing to ensure UI reliability and robust state management.
- ●Collaborated across teams to deliver production-ready features within agile timelines.
System Design Details
Architecture Overview
- ●React frontend interacting with REST APIs.
Key Engineering Decisions
- ●Focused on reusable components and test coverage during rapid development.
Challenges & Solutions
- ●Coordinating feature delivery across multiple teams and timelines.
Problem Solving
Get in touch.
I’m always happy to talk about backend systems, architecture, and performance engineering.
Typically respond within 24–48 hours