adarshxsh.
Portfolio · 2026

I turn difficult systems into products people can use.

I'm Adarsh. I build privacy-first mobile systems, distributed runtimes, and local intelligence tools—from the research question to the product surface.

Technical archive
Product systemslive
Systems Engineer Workspace
Built around
mobile · edge · local-first
Scroll to explore
Project journey

Nine chapters. One growing point of view.

Start with what I'm building now, then trace back through the systems and product ideas that shaped it.

01Now building · decentralized delivery

CIPHER

Open source
P2P infrastructure
The friction
File delivery should not depend on a centralized intermediary or a friendly network.
The construct
A Go and libp2p content network with encrypted chunks, Merkle verification, and NAT traversal.
The reality
Open-source transport for direct, verifiable exchange across difficult networks.
Golibp2pcryptography
Explore the technical case study
cipher / system viewlive
peer A
relay
peer B
02AttentionOS · on-device intelligence

SCOPE (Smart Contextual Opportunity & Priority Engine)

Mobile intelligence
The friction
Important notifications disappear inside a stream of promotional and social noise.
The construct
SCOPE triages Android notifications locally with a Kotlin listener and quantized TensorFlow Lite model.
The reality
A privacy-first attention layer that acts in under 5ms without exporting personal notification data.
TensorFlow LiteKotlinFlutter
Explore the technical case study
SCOPE / local triagelive
03Local developer intelligence

DevTrails Utility

Developer tooling
The friction
The research trail behind a development session is valuable, but highly personal.
The construct
An offline pipeline connects browsing history, local ML classification, code changes, and RAG-ready data.
The reality
Turns private research trails into useful developer context without exporting browsing history.
Pythonlocal MLRAG
Explore the technical case study
DEVTRAILS / intelligencelive
04HackSprint IIT Jodhpur · Team ShadowDevs

FATE (Full Attention Telemetry Engine)

2nd place
Digital wellness
The friction
Attention is shaped by context—not just screen time—and most tools ignore that reality.
The construct
FATE combines Android usage, place, and nearby-peer signals into a transparent attention score.
The reality
2nd place at HackSprint IIT Jodhpur, built with teammates for the Focus & Digital Wellness challenge.
Android nativetelemetryFastAPI
Explore the technical case study
FATE / attention contextlive
05A connected system reveal

MILF (Pocket Cloud)

Edge runtime
The friction
Mobile devices have compute power, but no cohesive way to safely turn it into a distributed runtime.
The construct
MILF connects a WASM consumer node, developer control plane, and orchestration layer into one platform.
The reality
A modular mobile infrastructure for lambdas and files—designed as one complete system.
WASMWAMRdistributed systems
Explore the technical case study
MILF / platform maplive
06MILF subsystem · execution edge

MILF Consumer Node (WAMR)

Mobile edge runtime
The friction
Executing untrusted workloads on a phone must be isolated, observable, and practical.
The construct
The Consumer Node embeds WAMR inside a sandboxed Android runtime for controlled WASM execution.
The reality
The execution layer that makes the wider MILF platform possible on mobile hardware.
WAMRAndroidWASI
Explore the technical case study
milf-wamr / system viewlive
wasm module
sandbox
Android
07MILF subsystem · control plane

MILF Dashboard

Developer experience
The friction
A distributed runtime is only useful when developers can inspect, control, and trust it.
The construct
A realtime dashboard brings deployment, execution, files, and observability into one surface.
The reality
The control plane that makes the MILF system legible and operable.
ReactWebSocketsobservability
Explore the technical case study
MILF / dashboardlive
08Social coordination on the move

LazyOne Mobile App (Flutter)

Mobile product
The friction
A request for help loses momentum when coordination breaks down away from the desk.
The construct
An offline-first Flutter client makes trust-aware task propagation available on mobile.
The reality
A resilient product surface for creating, discovering, and acting on tasks in motion.
Flutteroffline-firsttrust
Explore the technical case study
LAZYONE / mobilelive
09Where the journey began

LazyOne Web

1st place
Social task ecosystem
The friction
Task marketplaces optimize for transactions, not trusted social momentum.
The construct
LazyOne models task sharing as waves that move through a realtime trust graph.
The reality
1st place at a hackathon—the original product idea behind the wider LazyOne ecosystem.
Flutter webFirebasesocial graph
Explore the technical case study
LAZYONE / trust graphlive
Capabilities

Tools are useful when they unlock a real capability.

Mobile systems
Flutter · Kotlin · Android native · offline-first
Distributed & edge
Go · WASM · WAMR · networking · orchestration
Product engineering
React · TypeScript · realtime APIs · observability
Data & Intelligence
TensorFlow Lite · local ML · RAG · Python
How I build

Technical depth should make a product clearer, not colder.

01

Design for the constraints of real devices and networks.

02

Keep personal data local whenever the problem allows it.

03

Make complex systems understandable through the interface.

04

Build experiments into product surfaces that can be tested and improved.