Case Studies

Situation

SMEs struggled with fragmented tools across inventory management, order tracking, dispatch workflows, and team collaboration. Data lived in spreadsheets, messaging apps, and paper notes—creating inefficiency and errors.

Approach

• →Built unified iOS/iPadOS app using SwiftUI with offline-first Core Data + CloudKit sync
• →Implemented private + shared stores with NSPersistentCloudKitContainer for multi-user collaboration
• →Designed role-based access control and dispatch workflows with real-time status updates
• →Optimized for iPad with split-view interfaces for warehouse and office scenarios

Outcome

Single source of truth for inventory, orders, and dispatch. Teams collaborate seamlessly with offline capability and automatic sync. Reduced manual data entry and eliminated communication gaps between warehouse and office staff.

Situation

People struggle to maintain healthy habits and track fitness progress. Generic apps overwhelm users with features they don't need, while others lack the intelligence to provide meaningful insights or suggestions that adapt to user behavior.

Approach

• →Created clean SwiftUI interface focused on hobby tracking with visual streak indicators
• →Integrated HealthKit for seamless sync with Apple Health data (workouts, sleep, nutrition)
• →Leveraged Apple Intelligence Foundation Models for contextual habit suggestions
• →Designed privacy-first architecture with all data stored locally and synced via iCloud

Outcome

Users maintain consistent habits with AI-powered suggestions that feel natural and timely. HealthKit integration eliminates manual entry for fitness data. Streak visualizations provide motivation without gamification overload.

Situation

Keto diet followers face over-complicated nutrition tools designed for general audiences. Macro tracking requires too many taps, meal logging feels tedious, and most apps lack keto-specific guidance or quick-access complications for Apple Watch.

Approach

• →Built streamlined SwiftUI interface optimized for keto macro ratios (fat, protein, net carbs)
• →Integrated AI meal suggestions based on remaining macros and user preferences
• →Created Apple Watch complications for at-a-glance macro status throughout the day
• →Designed quick-log shortcuts and voice input for frictionless meal entry

Outcome

Users track macros effortlessly with keto-focused UI that eliminates unnecessary complexity. AI suggestions help plan meals that fit remaining daily macros. Watch complications provide instant feedback without opening the app.

Situation

Data analysts and scientists waste hours debugging messy CSV/Excel files with inconsistent dtypes, missing values, duplicate rows, and structural issues. Manual inspection in pandas is tedious and error-prone.

Approach

• →Built Flask + Pandas backend for automated validation of structure, dtypes, and data quality
• →Created clean Next.js frontend with drag-and-drop upload and visual issue highlighting
• →Implemented checks for common issues: missing values, duplicates, outliers, encoding problems
• →Designed downloadable reports with actionable suggestions for data cleaning

Outcome

Users identify dataset issues in seconds instead of hours. Visual feedback highlights problematic columns and rows. Automated suggestions reduce manual debugging and prevent downstream errors in analysis pipelines.

Situation

Traditional OS workflows weren't designed for AI-native usage. Users face friction integrating LLMs into daily tasks, privacy concerns with cloud-based AI, and lack of tools optimized for natural language command interfaces.

Approach

• →Developed experimental shell (Blossom Shell) for natural language command translation
• →Integrated local LLM inference with Ollama/llama.cpp for privacy-first AI processing
• →Explored context-aware assistants that understand user intent from conversational input
• →Built prototypes for AI-powered file management, search, and workflow automation

Outcome

Proof-of-concept tools demonstrating feasibility of AI-native OS interactions. Local inference ensures privacy while maintaining responsiveness. Natural language interfaces reduce cognitive load for complex tasks.