Pravat
Kumar Sahoo
I design distributed systems and own them from the first whiteboard to production.
13 years in, mostly in the enterprise space — Oracle, IBM, and now contracting through StarForge Ventures. Certified on AWS, Azure, and GCP, though the certs matter less than the systems still running.
Background
I'm Pravat. I've spent the last 13 years building backend systems — mostly in the enterprise space, at Oracle and IBM before moving into consulting via StarForge Ventures LLP. Right now I'm contracting as an Architect at Kickdrum, helping US-based clients ship AI-powered platforms.
My work tends to involve the full arc: initial discovery, architecture decisions, hands-on implementation, and making sure the thing actually lands in production. I've worked across talent platforms, real-time collaboration tools, aviation operations, and multi-region infrastructure pipelines — always owning more than just the code.
Outside of that, I mentor engineers when I can and try to write documentation people will actually read. I'm also a drummer and a cricketer, though those two don't often come up in the same conversation.
What I work with
Across the cloud stack, from infrastructure to AI integration to real-time distributed systems — these are the areas I've spent real time in.
Designing resilient, cost-optimised systems across AWS, Azure, and GCP. Certified professional on all three platforms.
Building production AI systems — LangChain pipelines, OCR-powered data extraction, agentic workflows, and LLM evaluation.
Microservices, event-driven architectures, real-time collaboration systems, and high-availability patterns at enterprise scale.
Java and Spring Boot at the core, with Node.js, Python, and Kotlin completing the stack. Strong, opinionated approach to clean APIs.
Automating infra across multi-region environments — business continuity, backup, and zero-drift deployments that ops teams trust.
End-to-end data engineering with lineage tracking, OCR extraction, and enterprise-grade quality gates across portfolio systems.
Projects
A few things I've worked on recently. Each one involved owning the architecture and staying involved through to delivery.
AI Talent Acquisition Platform
End-to-end AI-powered platform for a US-based PE company to streamline executive leadership hiring. LLM-assisted candidate matching with structured workflows and enterprise-grade access controls.
AI-Powered AWS Data Pipeline
AWS-native intelligent data pipeline for a PE firm — extracting and cleaning customer, contract, address, and subscription data across portfolio companies using OCR and LLMs, with full data lineage tracking.
Real-time Collaboration Platform
Collaborative UX design platform with real-time co-editing via CRDT (YJS) and integrated video/audio through AWS Chime. Built for concurrent multi-user workflows with conflict-free state management.
Business Continuity & IaC Automation
Multi-region infrastructure automation for an EdTech company — comprehensive storage and compute backup architecture, zero-drift infra management, and Terraform-based deployment pipelines.
Aviation Disruption Management System
Real-time disruption management for the aviation sector — handling operational alerts, automated rebooking workflows, and operational continuity under peak-load conditions using a fully serverless AWS stack.
Certifications & Recognition
Passed exams across AWS, Azure, and GCP over a couple of years. The scores are there too, if that matters to you.
Thinking Out Loud
Technical writing for engineers — architecture decisions, hard-won lessons, and opinions on building real systems.
On YouTube
Deep-dives, walkthroughs, and architectural breakdowns — for developers who learn by watching systems being built.
Channel Launching Soon
A series on cloud architecture, AI integrations, and engineering decision-making for senior developers and architects. Real problems, real solutions, no fluff. Think architecture walkthroughs, IaC deep-dives, and honest takes on building with LLMs.
Get notified at launch →Get in touch
If you have a project you'd like to discuss, or just want to talk through something technical — feel free to reach out.
Find me on the internet
I'm most responsive on LinkedIn. For consulting enquiries, email works best. Response time: usually within 24 hours.