Realtime Collaboration Board
A multiplayer whiteboard with live cursors, presence and conflict-free syncing, built for distributed teams that brainstorm together.
Open to new opportunities
Software Engineer & Researcher
I design and build reliable, user-focused software and explore research ideas that make technology more useful — from clean interfaces to the systems that power them.
About me
Navneet
I’m a software engineer who enjoys turning complex problems into simple, dependable products. I care about clean architecture, thoughtful user experience and code that’s a pleasure to maintain.
My work sits where engineering meets research: I like digging into how things work, prototyping ideas quickly and hardening the best ones into systems people rely on. Outside of work I contribute to open source, read papers and keep learning new tools.
Projects
A selection of projects I’ve designed and built — from production web apps to research prototypes and open-source tools.
A multiplayer whiteboard with live cursors, presence and conflict-free syncing, built for distributed teams that brainstorm together.
An extensible toolkit for benchmarking language-model pipelines, with reproducible experiment configs and a dashboard for comparing runs.
A fault-tolerant job queue with retries, priorities and dead-letter handling, designed to scale horizontally across workers.
Semantic search across research papers using embeddings, plus interactive citation graphs that make related work easier to discover.
A fast command-line tool that automates repetitive project chores such as scaffolding, release notes and environment checks.
The site you’re on: a fast, accessible single-page portfolio with dark mode, scroll-spy navigation and a working contact form — no framework needed.
Research
I enjoy research that bridges theory and practice — ideas that hold up both in the lab and in production.
We study parameter-efficient methods for adapting large language models to specialised domains with limited labelled data, and show that careful adapter design recovers most of the quality of full fine-tuning at a fraction of the training cost.
We present a scheduler that places mixed CPU and GPU workloads across heterogeneous clusters, improving utilisation and job completion times on traces from a shared research cluster.
We introduce a benchmark of realistic perturbations — noisy documents, conflicting sources and adversarial queries — and use it to measure how retrieval-augmented generation systems fail, along with simple mitigations.
Experience
Teams I’ve been part of and what I worked on — building products, running experiments and learning along the way.
– Present
Company Name · City, Country
–
University Lab Name · City, Country
–
Company Name · Remote
–
University Name
Focus on machine learning and distributed systems. Thesis on efficient model adaptation.
–
University Name
Coursework in algorithms, operating systems, databases and networks.
Skills
The languages, frameworks and tools I reach for to bring ideas to life.
Contact
Have an opportunity, a research collaboration or just want to say hi? Send me a message — I’ll get back to you as soon as I can.