Led the development of a web-based platform for UP Rural High School to replace manual grade encoding and class registration. Coordinated a 20-person development team and worked directly with school administrators and teachers to refine requirements and deliver a secure, real-time system.

Key Contributions:

• Stakeholder communication & requirements gathering
• Sprint planning, stand-ups, retrospectives
• Cross-functional team coordination

Developed new features across risk and data, user management, eKYC, and app analytics teams. Implemented Agile/Scrum workflows and CI/CD pipelines, reducing deployment time by ~25%.

Coordinated sprint planning and team organization for an Agile team developing a simulated autonomous lawnmower with obstacle avoidance and safety shutoff behaviors. Designed simple UI flows to help non-technical stakeholders visualize system behavior and progress.

Designed and implemented a Python-based virtual machine that simulates OS-level process execution, CPU scheduling, and memory management. The VM executes custom assembly-like programs at instruction-level granularity and supports multiple scheduling algorithms with detailed performance visualization.

Key Contributions:

• Implemented a virtual CPU executing one instruction per tick with full instruction decode and execution pipeline.
• Designed a flat byte-addressable memory model and extended it with a paging-based MMU.
• Built full process management using PCBs tracking PID, state, registers, program counter, execution time, and wait time.
• Implemented and validated multiple CPU scheduling algorithms: — First-Come First-Serve (FCFS) — Round Robin (preemptive, configurable quantum) — Multi-Level Feedback Queue (MFQ) with aging, promotion, and demotion
• Implemented paging and page-fault tracking, including logical address spaces, frame allocation, eviction, and MMU integration.
• Generated Gantt charts and performance visualizations to analyze scheduling behavior, wait times, and page fault trends across workloads.
• Built a shell interface for loading programs, running simulations, and debugging execution (core dumps, error dumps).

Designed and implemented a modular embedded system on an STM32F429ZI microcontroller, integrating multiple real-time subsystems including automated locking, lighting control, HVAC, fire alarm, intrusion detection, and an automatic front door. Built a hierarchical menu interface on an I2C LCD with keypad-driven navigation for user interaction.

Technical Highlights:

• Wrote interrupt-driven input handling (EXTI) for smoke, vibration, and PIR motion sensors.
• Configured multiple timers: — TIM1 for ultrasonic echo timing (input capture) — TIM3/TIM4 for PWM (servo motors, buzzer) — TIM6/TIM7 for periodic tasks and µs-level trigger pulses
• Designed state machines for fire alarm, security alarm, and door automation with priority resolution (fire > intrusion).
• Developed parameterized subsystems: lock scheduling, motion-activated lighting with timeout, HVAC mode switching.
• Implemented debouncing, asynchronous alert popups, and safe fallback states for partial sensor failures.
• Added hardware workarounds (LED door indicator, manual fire trigger) to maintain full system functionality with limited components.

Video demo and code available upon request.

Browser-based Python environment built for classroom use — no install required, designed for students to run code and see results immediately.

Built a Google Apps Script tool for a small appliance service business to improve technician scheduling. The add-on reads technician availability from Google Calendar and uses Google Maps drive-time estimates to suggest the best time slots for new jobs and generate an optimized stop order to reduce daily travel time.

Key features:

• Technician dropdown mapped to individual Google Calendars
• “Best slot” suggestions based on open gaps and estimated drive time
• Route optimization to reduce total daily travel
• Input validation and caching to improve reliability and speed

A lesson-planning dashboard built for coding coaches at theCoderSchool. CoachNotes centralized session prep, post-session notes, and parent-facing write-ups in one place — replacing a fragmented workflow of spreadsheets and manual copy-paste.

What I built:

• Full-stack web app on Google Apps Script with a Google Sheets backend, served as a container-bound web app with zero external dependencies
• Student roster with session history, scheduled day/time, and language tracking across 100+ cumulative sessions per student
• Lesson prep system with per-session “Option 1 / Option 2” cards (label, overview, concepts, sample notes, sample code) generated manually or via Claude AI integration
• AI-assisted prep workflow: one-click export of session history as a structured prompt → paste into Claude → import JSON response back as formatted prep options
• Pike13 integration: full session history import via copy-paste parsing, and a post-session note sync that matches note blocks to students by name and writes back language, project, concepts, and status
• Day view showing all sessions for a selected date, collapsible per-student cards, and bulk session management

What I learned: Shipped a functional internal tool used in active coaching sessions. The project was ultimately not adopted beyond my own use — theCoderSchool franchises follow a centralized tooling recommendation (Notes+), and individual coaches don’t have purchasing authority. That constraint surfaced a real product lesson: for B2B tools in franchise environments, the decision-maker and the end user are rarely the same person. The technical and UX work was validated; the go-to-market assumption wasn’t.

Built a self-contained browser tool that converts Strong app workout exports into Garmin .fit files, preserving full strength training detail — exercises, sets, reps, and weights — for accurate Strava uploads. Solved import bugs around UTC timestamp anchoring, exercise grouping by category/subtype pairing, and FIT’s 1/16-kg weight resolution to avoid undercounted lifts. Also includes a Python-based bulk converter for full CSV exports.