Harmanpreet Singh, software engineer
My latest project, HomeSense, builds a platform around my thesis model for smart-home activity recognition: live streaming, anomaly review, web and phone apps, a resident simulator and a Kubernetes deployment. Before that I shipped production web applications at Stylabs Technologies.
Synthetic day, simulated resident
Events
Invented data for illustration, not a recording of a real home. Now: 06:59:54.
Latest project, June to October 2026
I took my thesis model for smart-home activity recognition and built the platform a real deployment would need around it: a streaming pipeline, anomaly review, a controlled assistant, a resident simulator, gated model releases, web and phone apps, and Kubernetes packaging.
Sensors
Kafka
Window builder
Predictor
Anomaly detectors
PostgreSQL
Web and phone
27
architecture decision records
One per major design choice, written as it was made.
499
Python test functions
Unit, contract, golden-number and integration suites.
15
workloads in one Helm chart
Each with probes, autoscaling and a network policy.
47,211
events stored out of 47,211 published
After deleting stream pods mid-simulation on a local three-node cluster: none lost, none duplicated.
Screenshots are real captures from the project's UI guide, taken from a seeded demo house built on the public CASAS Aruba floor plan. Figures come from the project's own tests, benchmarks and recovery checks, run on a development workstation and a local cluster.
Professional work at Stylabs Technologies
Products I worked on that are live today. The screenshots are each product's public home page and show the whole product, so each card says which part was mine. The code belongs to the company.
Live
Holiday-home rental product
A live product for design-led holiday homes and service apartments in Dubai, for daily and monthly stays. I built the chat support system for customers, employees and owners, and contributed rental, sales and expense features.
Live
Villa rental product
A live product for renting luxury villas and holiday homes in India, searchable by location, dates and guests. A sister product to Manzil, where I worked on the same chat support system and rental, sales and expense features.
Live
Production web platform
A live Dubai real-estate platform for buying, renting and listing property, built by a small team on a tight timeline. I helped set up the foundation and built frontend components, backend integration, authentication and data workflows.
Live
E-commerce marketplace
A live online marketplace for sneakers, streetwear, apparel, watches and more. I developed components, fixed bugs, built API integrations, tested the application and wrote API documentation for the frontend and mobile teams.
M.S. thesis, University of Georgia, 2026
Multimodal Self-Supervised Activity Recognition on CASAS Smart-Home Sensor Streams
A smart home can tell what a resident is doing without a camera, just from motion and door sensors clicking on and off. My thesis asks how best to read those sparse event logs, and finds that the answer depends on the house.
58 pages, 2.2 MB. Advisor: Fei Dou.
4
CASAS homes
Aruba, Cairo, Kyoto and Milan
4
views of each window
sequence, text, image and graph
30
events per window
stride of 15, so windows overlap
3
random seeds
every result is a mean over seeds 10, 30 and 50
Submitted to a conference, under review
FACET: Factorized Ambient-Sensor Experts with Local Competence-Weighted Fusion for Smart-Home Activity Recognition
A paper on recognising everyday activities from ambient smart-home sensors, submitted to a conference and currently under review.
Joint work: I built this together with a collaborator.
Work, coursework and background
Research, professional team work and systems exercises, each with my own contribution stated.
Independent project: Simulation and learning tool
An interactive way to learn system design: instead of reading about caching, you flush the cache and watch the database fall over. A simulation engine drives 40 system labs, an architecture canvas, a code playground and scored challenges. It has helped over 10,000 users learn system design.
11 projects from my M.S. courses, each rebuilt as a working demo in the browser.
Independent project: Interactive learning tool
A browser-only site that teaches machine learning by letting you change it: 75 lessons with live interactive labs and 20 guided end-to-end projects. Every model is written from scratch in TypeScript, so there is no server and nothing to install. It has helped over 5,000 users learn machine learning algorithms.
Independent project: Real-time web and mobile app
A watch-party app for people who follow sports together: spoiler-safe chat, live reaction counts, polls, pick'ems and a data desk with live tables and fixtures for UFC, F1, football, college football and cricket. A web app, a Node server and an Android app share one API. Around 200 users have used it.
What I work on
Frontend components, API integration, authentication and data workflows in team-built web products.
Evidence: HomeSense platform
What I work on
Turning irregular sensor logs into model inputs, comparing models fairly, and reporting results with their metric and context.
Evidence: Smart-home activity recognition
What I work on
Streaming, idempotent processing and orchestration in HomeSense, plus sockets, concurrency, hashing and query processing built from scratch in coursework.
Evidence: HomeSense platform
Experience
Technical Intern, full-stack, May 2023 to May 2024
Contributed to several production web applications on a product team, working across the client, the API and the database.
Live products I worked on
Education
M.S. Computer Science (thesis track), Expected December 2026
Graduate coursework
Education
B.E. Computer Engineering, 2019 to 2023
First Class with Distinction
Algorithms and theory
Systems, networks and data
AI and machine learning
Software engineering
About
I started in software engineering at Stylabs Technologies, where I learned to ship features inside team codebases. Graduate study at the University of Georgia added systems, databases and machine learning, and my thesis combined them. HomeSense is where I took that all the way to a running platform.
Email is the quickest way to reach me. I am also on GitHub and LinkedIn.