About
9+ years of shipping. Now building what's next.
I'm Pragash — a full-stack engineer who has spent close to a decade turning ambiguous product problems into reliable software. Actively interviewing, and available to start immediately.
The journey
My career started in 2017 at Relevantz, working on a high-traffic bidding and EMI payments platform — the kind of system where a slow page directly costs money. I learned performance engineering the honest way: lazy loading, memoization, and a legacy AngularJS-to-Angular migration that improved runtime performance by 40% while the site kept serving customers.
At RexEMR I got to build a healthcare SaaS platform end-to-end — component library, REST APIs, and the insurance claims workflows where a validation bug means a clinician doesn't get paid. Payment-critical software teaches you a respect for edge cases that no tutorial can.
From 2022 through July 2026 I worked with Dell Technologies (through Turing) as a senior consultant: micro frontends aligned to enterprise design systems, .NET Core and Node.js microservices behind an Ocelot API gateway, real-time systems on SignalR sustaining ~1,200 concurrent users. Fully remote, fully async, across time zones — which is where I learned that written communication is an engineering skill.
That engagement wrapped in July 2026, and I made a deliberate choice about what came next: rather than taking the first available contract, I went full time into applied AI. Since July I've been building the systems instead of reading about them — a retrieval pipeline and agent loop from first principles, a spec-driven quant trading platform, time-series analytics in kdb+/q — while finishing my MBA. So I come into this search with 9+ years of production discipline and hands-on AI systems work, not one or the other.
The turn toward AI
I adopted AI-assisted development early and watched it change what one engineer can ship. That led to a deeper question: how do these systems actually work? So I started building them — a RAG pipeline and agent loop from first principles (embeddings, ChromaDB, tool calling, local LLMs via Ollama), a spec-driven quant trading platform, and time-series analytics in kdb+/q. Not courses. Working systems.
My conviction: the most valuable engineers in the AI era are the ones who combine production engineering discipline — observability, API design, failure handling — with a real understanding of how LLM systems behave. That intersection is exactly where I'm building.
What I'm looking for
I'm actively interviewing and available to start immediately. The fit I want: a senior full-stack, AI, or founding engineer role on a product-driven, remote-first team where I own real surface area — ideally somewhere LLM and retrieval systems are part of the product, not a side experiment. Remote worldwide, and open to relocating to the UAE. If that sounds like your team, get in touch — I reply within a day.
How I work
Spec first, code second
Every serious system I build starts with a written design — my quant platform began as a SYSTEM_DESIGN.md before a single line of Python. Specs surface the hard decisions early, make AI-assisted coding dramatically more effective, and leave a contract the code can be reviewed against.
Impact over activity
Shipping features is easy; moving numbers is the job. The work I'm proudest of reads as outcomes: claims rejections down 20%, initial render down 35%, 6–8 hours of cross-team troubleshooting saved per week.
AI as a force multiplier
I've used AI-assisted development in production since 2022 — not to skip thinking, but to compress the distance between a clear spec and working code. The engineers who thrive next are the ones who can direct AI with precision, then verify rigorously.
Understand the layer below
I don't use RAG without building an embedding pipeline by hand, and I didn't trust ORMs before writing the SQL. Building one layer deeper than you operate is the cheapest insurance in engineering.