RAG without the theater
Chunking, search, and citations beat demo polish. Notes on document QA that still works after the meeting ends.
ReadAbout
I approach problems through systems architecture and product reasoning. I focus on how technical decisions affect user outcomes, not just whether something can be built. My M.Tech coursework at the National University of Singapore emphasizes architectural reasoning—evaluating trade-offs, scalability, and maintainability—while my exposure to product management and design has shaped how I think about user needs and constraints.
Systems architecture & trade-offs
Product reasoning & user needs
AI systems & RAG pipelines
Architectural decision-making
Engineering decisions grounded in trade-offs
Focus & Strengths
Projects & Exposure
Education & Foundations
How I think about systems
I focus on systems thinking and architectural reasoning. My strength lies in understanding how technical decisions impact user outcomes and system behavior over time. I evaluate trade-offs, scalability, and maintainability as first-class concerns, not afterthoughts.
I bring product reasoning to technical work. Rather than building in isolation, I consider user needs, constraints, and long-term implications.
My exposure to product management and design has shaped how I think about why we build what we build, not just how.
I have foundations in AI systems and RAG architecture. I understand how retrieval pipelines, vector search, and language models integrate into larger systems. I design systems where AI components are replaceable and well-architected, not just functional.
Where I applied this thinking
SpaceFlow — Smart Workplace Platform
A system design exploration demonstrating how product needs shape architectural decisions. I designed a modular, service-oriented platform for space booking, occupancy tracking, and analytics.
The project focused on service boundaries, API-first design, and how different concerns (booking, analytics, AI insights) work together without tight coupling.
DocuMind (Insight Garden)
A full-stack document RAG SaaS: users upload PDFs, the system chunks and embeds them into a vector database, and users chat with their documents via streaming AI—with source attribution and configurable LLM/embedding providers.
The project emphasized end-to-end product design: auth, async processing with progress, hybrid retrieval (pgvector + lexical), SSE streaming, and deployment (Vercel, Railway, Supabase, Upstash)—with AI capabilities as replaceable components.
Plant Disease Recognition System
An end-to-end ML system that classifies plant leaf images into 38 disease/healthy classes using a CNN. Single-source config and data pipeline keep training and inference aligned; experiment logging and a Streamlit web app round out the project.
I designed the full pipeline: centralized config.py and
data_pipeline.py, TensorFlow/Keras training with EarlyStopping and
experiment logs, and a thin Streamlit app with correct preprocessing and model
caching—plus CI (GitHub Actions) and deployment docs for Streamlit Community Cloud.
These projects demonstrate decision-making in system design. I focused on trade-offs, system boundaries, and architectural reasoning—not just implementing features.
My technical foundations
M.Tech (Software Engineering)
National University of Singapore | GPA: 4/5
My coursework emphasizes architectural reasoning, systems design, and evaluating technical trade-offs. I've studied how to design scalable systems, reason about service boundaries, and make decisions that balance technical depth with practical constraints.
I have exposure to product management and product design through coursework and projects. This has shaped how I think about connecting technical decisions to user needs and business outcomes.
Systems architecture
AI systems
Data pipelines
Product reasoning
Architectural decision-making
B.Tech (Information Technology)
Vellore Institute of Technology, Vellore, Tamil Nadu | CGPA: 8.21/10
My undergraduate studies provided a strong foundation in computer science fundamentals, software development, and information systems. Core programming concepts, database design, web technologies, and software engineering principles form the basis of my technical approach.
My Certifications
Designing Modern Software Systems
NUS-ISS. Graduate Certificate covering system design principles, modern software architecture, and evaluating trade-offs, scalability, and maintainability.
ViewArchitecting Scalable Systems
NUS-ISS. Graduate Certificate covering scalable distributed systems, architectural patterns, and reliability and performance considerations.
ViewAWS Cloud Practitioner
AWS. Core cloud services, security fundamentals, and reasoning about cloud architecture and cost optimization.
ViewMicrosoft Azure AZ-104
Microsoft. Azure administration, resource management, and designing and deploying cloud infrastructure.
ViewGenerative AI
IBM. Generative AI models, prompt engineering, and integrating AI capabilities into applications.
ViewAgile with Atlassian Jira
Atlassian. Agile methodologies, sprint planning, and using Jira for project management and team collaboration.
ViewNUS ISS Learning Festival 2025
NUS Institute of Systems Science. Explored AI systems, cloud architecture, and digital transformation strategies in enterprise contexts.
ViewSelected Projects
DocuMind (Insight Garden)
Full-stack document RAG SaaS: upload PDFs, analyse them with AI, and chat with your documents via streaming answers—with source attribution. React, NestJS, pgvector, BullMQ. Deployed on Vercel, Railway, Supabase, Upstash.
Check demo View Case Study
Plant Disease Recognition System
End-to-end ML pipeline: CNN classifies plant leaf images into 38 disease/healthy classes (~95% accuracy). TensorFlow, Keras, single config & data pipeline, Streamlit app, CI/CD.
Check demo View Case Study
EasyMeals
A concept case study exploring subscription-based healthy meal delivery for busy professionals. Product strategy and design thinking.
View Case Study
SpaceFlow — Smart Workplace Platform
A system design exploration of space booking, occupancy analytics, and AI-assisted insights.
View Case StudyBlog
Why we build what we build
Name the problem before the stack debate runs away. How I try to keep intent and architecture pointed at the same thing.
Read
What to build is still the hard part
Lots of noise about tools and prompts. Almost none about which problem deserves your month. Where I am putting attention.
Read
Claude and agentic building
Tight loops with Claude over your repo and terminal. What helped, what lied, what I still own.
Read