Eolus AI
An intelligent document analysis platform leveraging RAG pipelines, vector search, and multi-modal LLMs to surface insight from unstructured data at scale.
View projectBuilding AI systems that solve real problems through intelligent software, thoughtful engineering, and modern design.
Every system I build starts with a clear problem, an honest understanding of constraints, and the commitment to get it right — not just working, but right.
[ Explore My Work ]Engineering is not just about writing code — it is about making decisions under uncertainty, designing systems that outlast their requirements, and understanding when to build and when to wait.
I combine deep AI research fluency with full-stack engineering to build products that are precise, scalable, and thoughtfully designed. Not just working software, but the right software.
What unites these domains isn't a single technology — it's a clear thesis: intelligent systems should be useful, reliable, and designed with intention.
Built with rigor. Designed with purpose.
Each project is a deliberate engineering decision.
An intelligent document analysis platform leveraging RAG pipelines, vector search, and multi-modal LLMs to surface insight from unstructured data at scale.
View projectBuilt and evaluated a neural network to classify potential exoplanets from NASA Kepler light-curve data — with visualised training metrics and end-to-end model evaluation pipelines.
View on GitHubResponsive web application comparing hotel and flight prices across platforms via third-party REST APIs — optimised for fast search results and a clean, frictionless UX.
View on GitHubGesture-based AR music controller using MediaPipe Hands and TensorFlow.js — enabling touch-free, real-time hand-gesture control of audio playback entirely in-browser.
View on GitHubBuilt at the intersection of language intelligence and enterprise-grade engineering. Eolus ingests raw, unstructured documents — contracts, research papers, internal wikis — and transforms them into a queryable, reasoning layer for organisations.
The architecture combines dense vector retrieval, BM25 hybrid search, and a fine-tuned generation layer — designed to be accurate, fast, and auditable.
"The most important engineering decisions are the ones you make before writing a single line of code."Daanveer Mehra
Serious AI work requires more than model calls and API wrappers. It requires systems thinking, engineering judgment, and the patience to get things right.
Every system I build begins with a clear understanding of the problem space. I read the literature, test assumptions, and design before I code.
Prototypes are hypotheses. Production systems are commitments. I design for reliability, observability, and graceful failure from day one.
Good design is a technical discipline. I treat interface decisions with the same rigor as architecture decisions — because both shape what gets built.
The best engineering decisions optimize for years, not sprints. I build systems that can evolve, adapt, and survive requirements changes.