Here are some of the projects I have developed and architected. Feel free to explore:
Engineering two interconnected AI products from the ground up as a solo founder. Soulspace is the world's first AI social network — a persistent social world populated by 2,500+ autonomous AI personas with unique identities, memories, and evolving relationships. It is designed to push the frontiers of human and AI interaction at a social level: not just one-on-one chat, but a living environment where humans and AI coexist, connect, and shape each other over time. Soulmate is a companion platform layered on top, featuring persistent memory, real-time voice conversations, and proactive messaging that reaches out to users — not the other way around. Built on Django, Gemini Live API, pgvector, Cloudflare R2, Cartesia TTS, and Cloud Run. The proprietary emotional conversational dataset being generated is designed to become a long-term AI research moat.
Solo-shipped a full-stack B2B job platform connecting Tanzanian safari and hospitality companies with prospective employees — Africa's first dedicated safari industry job board. Built with Next.js 16, React 19, TypeScript, Supabase/PostgreSQL, Tailwind CSS v4, and TanStack Query. Implemented Row Level Security (RLS), Supabase Auth, Zod schema validation, and rate-limited endpoints for a secure backend. Engineered a multi-tier employer portal with Role-Based Access Control (RBAC), a Kanban-style applicant tracking pipeline, PDF generation via @react-pdf/renderer, automated email flows via Resend, and M-Pesa localized payment integration. Optimized for low-bandwidth environments targeting sub-3-second load times on 3G networks through server-side rendering, skeleton loaders, and Supabase Storage image optimization.
Designed a civic reporting concept for Vancouver that reimagines how residents interact with local government. The product combines map-based issue intake, photo and voice uploads, AI-assisted report drafting, and public issue visibility — so residents can report problems in seconds and see what others in their neighbourhood have flagged. Built around structured reporting workflows, stakeholder impact visibility, and a dramatically better civic UX than existing 311 systems. Designed to reduce friction between citizens and city infrastructure teams.
Built a 73% accurate machine learning model to predict match winners, solving the critical challenge of data leakage by exclusively using in-game data from the first 15 minutes of play. Addressed multicollinearity in raw stats by engineering features like gold_difference and dragon_advantage to better represent team advantage. Selected a tuned Random Forest model to handle complex non-linear relationships between features. Interpreted feature importance scores to surface actionable insight — confirming early-game gold_difference as the strongest predictor of victory.
Developed and implemented a user-friendly web-based inventory management system for the textile manufacturing industry using Django and PostgreSQL, significantly improving efficiency in tracking raw materials and finished products. Designed real-time data synchronization ensuring accurate inventory information across departments. Collaborated with cross-functional teams to gather requirements, design system architecture, and integrate the solution into an existing workflow — resulting in a streamlined and automated inventory control process. Authored full documentation including user manuals to support adoption and long-term sustainability.
Designed and developed a convolutional neural network (CNN) model in Python using TensorFlow to classify handwritten digits. Successfully preprocessed the MNIST dataset — handling data loading, normalization, and augmentation — to ensure optimal model training. Overcame challenges in training and optimization, and maintained detailed code documentation including clear function descriptions to facilitate reproducibility.
Architected a comprehensive trading analytics platform designed to backtest strategies, log daily trades across forex and options markets, and provide actionable statistical insights based on historical trading data. Focused on surfacing patterns in trade performance to support disciplined decision-making.
Designed, developed, and deployed a fully responsive, accessible website for a non-profit organization based in Tanzania to establish their digital presence and streamline communication with donors and the communities they serve.
Implemented attention mechanisms in a pre-trained sequence-to-sequence (seq2seq) neural machine translation (NMT) model, significantly improving translation performance. Produced a comprehensive report summarizing the task, methodology, results, and proposed improvements.
Developed a machine learning model capable of detecting and classifying human emotions from facial images. Applied computer vision techniques and deep learning to build a real-time emotion recognition pipeline.
Replicated core e-commerce functionalities from the SHEIN platform including user authentication, dynamic product pages, a payment gateway simulation, and a cart with coupon functionality. Built using HTML, CSS, JavaScript, and LocalStorage for client-side state management.
Developed a Python script to validate game rules for games like Blackjack and Poker using the unittest library. Designed comprehensive test cases ensuring compliance with predefined rule requirements, and documented findings with proposed solutions for improved rule adherence.
A React-based weather application that fetches and displays real-time weather information using the Open Weather API, with responsive design for mobile and desktop.
Developed an advice generator app using HTML, CSS, JavaScript, and the Advice Slip API. Implemented responsive design for optimal layout across devices, dynamic advice generation on button click, and interactive hover states.
Built a Sudoku app for Android using Java and Kotlin, following Agile methodologies for project management. Utilized Git for version control and implemented external file storage to support dynamic puzzle list updates.
Tic Tac Toe desktop app built using Python and tkinter.
I'm Yash Serai — a builder working at the intersection of social networking and AI companionship. My flagship projects are Soulspace, the world's first AI social network featuring thousands of autonomous AI personas living in a persistent social world, and Soulmate, a companion platform built on top of it with persistent memory, real-time voice, and proactive messaging. Both products are in active development as a solo founder while I work full-time as a Data Engineer.
By day, I work as a Data Engineer at Opensignal, which acquired my previous employer ThinkCX. I was the only junior engineer retained through the acquisition — a fact I'm proud of. In that role I've built automated QA systems validating hundreds of business logic rules daily, engineered data pipelines that unlocked previously unmappable datasets, and shipped reporting infrastructure that saves clients 10+ hours of manual work per week.
Outside of my day job, I've shipped Safari Gigs — Africa's first dedicated safari industry job board — solo, handling everything from database schema and auth to M-Pesa payment integration and performance optimization for 3G networks in East Africa. That project sharpened my instincts for building under real-world constraints: low bandwidth, localized payments, and users who need the product to just work.
I hold a Bachelor of Science in Computing Science from Simon Fraser University. My background spans machine learning, full-stack development, DevOps, and data engineering. I'm drawn to problems that sit at the edge of what software can do — especially where AI, human connection, and scale intersect.
Languages: Python, JavaScript, TypeScript, SQL, C, C++, R, Java, Kotlin, Matlab, Go
Frontend: Next.js, React, React Native, Tailwind CSS, shadcn/ui, Radix UI, Framer Motion, TanStack Query, React Hook Form, Zod
Backend & Data: Django, Node.js, PostgreSQL, Supabase, pgvector, Celery, REST APIs, Firestore, Railway, Cloud Run, Cloudflare R2
AI/ML: TensorFlow, PyTorch, Scikit-learn, Gemini / Gemini Live API, Vertex AI, Cartesia TTS, Deepgram STT, Numpy, Pandas, Matplotlib
DevOps & Tools: Git, Docker, Linux, Bash, JUnit, SQLite, Microsoft Azure
Certifications: Microsoft Certified: Azure Developer Associate | Microsoft Certified: Azure AI Engineer Associate
Joined ThinkCX as a Data Engineer and was selected as the only junior engineer retained following Opensignal's acquisition of the company — all other junior data engineers were let go during the transition. Deployed a fully automated QA system in Python/SQL that validates daily data against over 700 business logic flags. Improved overall data accuracy by 27% by identifying and resolving a critical data integrity flaw in a key client's dataset, earning a personal commendation. Engineered a specialized data pipeline using caller ID to map and ingest data from previously unmappable network ports, expanding the usable dataset by 13%. Developed an automated reporting module that generates and distributes key data visualizations for clients, significantly improving client satisfaction and reducing manual work by 10+ hours per week.
Maintained a 95% five-star customer rating by delivering personalized product recommendations and engaging product demonstrations. Consistently adapted to diverse customer needs in a fast-paced retail environment while collaborating effectively with colleagues across departments.
Monitored and optimized server performance, leading to a 20% reduction in system downtimes and improving overall station operations. Developed detailed technical documentation for system configurations and troubleshooting procedures, significantly reducing resolution time for common issues and supporting efficient onboarding. Assisted in the setup and maintenance of network infrastructure including routers, switches, and firewalls, improving network security and data flow by 25%. Implemented automation scripts for routine maintenance tasks, reducing operational workload by 40%.
Identified and resolved software bugs contributing to a 15% increase in system stability, earning commendation for meticulous debugging and code optimization. Authored comprehensive documentation for the Inventory Management project, significantly reducing onboarding time for new team members. Assisted with the development and implementation of a real-time inventory tracking system resulting in a 30% reduction in stock discrepancies.