Notable Project Work

Streamphones automates multi phone management from a Windows PC using configuration driven device definitions and per device monitor loops. It handles WiFi drops, reboots, and scrcpy failures so a multi device setup remains usable without constant manual intervention.

Devices are defined in a simple text format: deviceName, ip:port.

Impact

This project turned a fragile manual setup into a stable dashboard style workflow where I can launch once and let the system self heal when connections drop.

Highlights

• Config driven device list and launch selection prompt
• Layout presets with predictable multi window positioning
• Per device monitor loop with scrcpy health checks
• ADB reconnect loop with network reachability checks
• Session IDs and lock files to prevent duplicate monitors
• Shutdown flag support for clean exits
• Optional brightness adjustment for unattended viewing

Tech stack: Windows batch scripting, ADB over TCP, scrcpy, PowerShell networking checks, configuration driven orchestration

Skills demonstrated

• Windows automation and process orchestration
• Reliability patterns with monitors, restart logic, and locks
• Network aware reconnect tooling for TCP device endpoints
• Configuration driven scaling for multi device workflows
• Practical ADB and scrcpy operations in real conditions

I developed task nodes for a game automation framework with reliable progression through interaction trees and UI prompts that often cause runs to stall. I prioritized continuing interactions when possible, then selecting options as a fallback. I also added post flow widget cleanup so scripts can continue without manual intervention.

Highlights

• State based task flow with clear transitions
• Interaction handling that checks continue before option selection
• Widget interaction for dismissing completion prompts
• Logging and debugging hooks for stuck states

Tech stack: Java, client API integration, UI widget interaction, state machines, logging

Resume bullets

• Implemented automation nodes with reliable interaction selection and continuation checks to prevent stall loops
• Added UI widget handling to close completion prompts automatically and resume navigation
• Improved script stability by introducing explicit state gating so preparation steps only run when required

I created a server side pipeline to ingest documentation and community data into PostgreSQL for development search workflows. I validated ingestion quality with SQL checks and identified rows where cleaned content was missing even when raw content existed, then used those results to guide pipeline fixes.

Highlights

• Ingestion validation with SQL counts and null checks
• Raw content and cleaned content field separation
• Focus on RAG retrieval quality for developer questions

Tech stack: Python, PostgreSQL, MCP tooling, SQL validation queries, ingestion pipelines

Resume bullets

• Built a Postgres backed knowledge store for documentation and community content to improve LLM assisted coding workflows
• Used SQL auditing queries to detect ingestion failures where cleaned text was missing and prioritized fixes
• Designed schema and pipeline steps to support future indexing and retrieval improvements

I wrote Cypher queries to match contacts by properties, create relationships based on shared area codes, query contacts by relationship membership, and remove relationships when requirements changed. The scripts were designed for repeatable execution with consistent results.

Highlights

• Relationship creation from property driven rules
• Combined relationship and property filtering queries
• Relationship deletion and rebuild workflows

Tech stack: Neo4j, Cypher

Resume bullets

• Authored Cypher scripts to create and query area code relationships among contact nodes
• Built queries that return node data based on relationship membership and attribute filters
• Implemented relationship deletion steps to keep the graph consistent as rules evolved

I set up and ran a heavily modded Minecraft server on Linux for the High Tech Cobblemon pack, managing a large mod list and the server environment needed to keep it running reliably. The work focused on practical server operations, compatibility management, and keeping performance acceptable for real players.

Skills demonstrated

• Linux server administration for game servers
• Modpack and dependency management at large scale
• Troubleshooting startup crashes and mod conflicts
• Performance minded hosting and stability practices
• Repeatable setup, config management, and maintenance

Highlights

• Tuned Java memory and startup options for large mod loads
• Used crash logs to resolve conflicts and stabilize launches
• Maintained restart, backup, and maintenance routines
• Configured networking and access controls for players

Tech stack: Linux, Minecraft server administration, Java runtime tuning, modpack operations, configuration management

I operate containerized workloads and virtual machines with a focus on repeatable deployment, permission safety, and service continuity. Work includes mount mapping, volume strategy, networking setup, and root cause analysis when services fail due to host level constraints.

Highlights

• Container lifecycle management with reproducible configs
• Volume and SMB mount strategy for persistent data
• Permission debugging across host and container boundaries
• VM management and service isolation for lab workloads

Tech stack: Docker, docker compose, Linux, Unraid, Proxmox, SMB mounts, volume permissions

I configured Sonarr with custom formats inspired by community standards and split dual audio detection into language pair formats. I used tags to reflect content reality across dual, partial, dub, and sub signals while avoiding conflicting tags when releases include mixed attributes.

Highlights

• Language pair custom formats for stronger classification
• Tag semantics aligned to actual audio and subtitle coverage
• Portable Docker based setup on Unraid

Tech stack: Sonarr, Docker, Unraid, custom formats, tagging rules

Resume bullets

• Designed language detection custom formats and tags to classify dual audio and sub only releases reliably
• Implemented tag logic that avoids false signals while still capturing mixed language edge cases
• Deployed and maintained the workflow in Docker on Unraid for consistent automation

I completed coursework projects covering conceptual modeling, ER diagrams, table creation with constraints, controlled primary key strategies, referential integrity testing, and complex queries. I also worked with Sakila style data for join, aggregation, and subquery assignments.

Highlights

• ER model to schema translation
• Referential integrity and null behavior testing
• Join heavy query sets with aggregation and subqueries

Tech stack: MySQL, SQL, relational modeling, Sakila dataset

Resume bullets

• Built normalized relational schemas from ER diagrams and validated data integrity with constraint tests
• Authored SQL query scripts using joins, aggregation, and subqueries to solve dataset driven prompts
• Tested edge conditions including null handling and referential integrity enforcement

I built a practical streaming setup where FFmpeg repeatedly overwrites a single JPEG frame and serves it through an MJPEG server. I tested ZeroMQ buffering adjustments for smoother playback and then planned a threaded decode approach to separate decoding from transport and reduce stalls.

FFmpeg writes frames at about 10 fps to /tmp/mjpeg_frames/frame.jpg, served via mjpg-streamer using the input_file plugin on port 8082.

Highlights

• MJPEG frame overwrite pattern for low overhead live preview
• ZeroMQ transport tuning for smoother playback
• Threading plan to isolate decode work and stabilize flow
• Targeted continuous detection with bounding box overlays

Tech stack: Python, FFmpeg, MJPEG streaming, ZeroMQ tuning, threaded design

Resume bullets

• Built a real time MJPEG streaming pipeline using FFmpeg frame generation and mjpg streamer for live preview
• Tuned transport buffering to improve smoothness and tracked artifact tradeoffs for the next iteration
• Planned a threaded decode architecture to reduce stalls and improve visual stability for overlay workflows

I developed a workflow to disable and re enable a specific HID device using pnputil after alternate disable approaches failed due to device criticality checks. This supported safe timed toggling as part of a larger automation pipeline.

Highlights

• CLI device control using exact hardware instance ids
• Method failure debugging and approach replacement
• Timed toggle support for macro pipeline control

Tech stack: Python, Windows, pnputil, HID device management

I built a sorted list ADT in Python with insertion, removal, search, iteration, and indexed access. I also implemented a validated Course ADT, debugged binary search tree edge cases, and built a custom HashMap ADT to cache recursive calculations for large speedups.

Highlights

• Input validation and defensive programming
• Custom iterability and index support
• Hash based caching to reduce redundant recursion work
• Debugging of tree edge cases

Tech stack: Python, data structures, recursion optimization, testing mindset

Resume bullets

• Built from scratch data structures with validation, iteration, and sorted behavior, then fixed edge case failures
• Implemented caching strategies that reduced repeated recursion work and improved runtime performance
• Diagnosed and corrected tree removal edge cases, including root replacement and traversal consistency