Jacob Lu
Hello!
​
I’m Jacob Lu, a junior at WPI, formerly a software engineer at Philips, and currently a game developer at All In Tactics. I build the systems that bring design ideas to life, from chess games, to large-scale testing and infrastructure in health monitoring products. Whether I'm working on a game or a piece of software, I always try to deliver a product that might help someone smile.
Game Design Projects
Cozy Chessu
Role: Systems Engineer
Contributions:
Worked mainly on designing + programming foundational game systems.
Pushed for and implemented player to NPC interaction, bringing the game's environment to life.
Built the business tycoon, narrative gameplay systems.​
Studio: All In Tactics
Platform: PC
Tools: Unreal Engine 5, C++
Dragonfly Engine
Role: Programmer
Contributions:
Collaborated on a two person team to create a 2D game engine from scratch and use that engine to program a sample game. Built core engine physics and iteration loop. Programmed sample game, integrating partner's art and sound design assets.
Course: Technical Game Development
Platform: PC
Tools: Visual Studio 2022, C++
Technical Projects
Parallelization Of Philips Testing Framework
Purpose: Eliminate test execution duration bottle neck that prevented nightly builds (new version of product generated) from finishing within 24 hours
​
Work: In order to update the product(a bedside health monitoring system), a large number of tests must be run using new code to see if functionality has broken. To reduce the execution time of this process, I built a parallelized testing framework which utilized multiple clients connected to a central server to run 'batches' of tests in syncronous.
​
Result: Test execution duration was significantly reduced by 66% and the bottleneck was eliminated.
Enhancement To Gen-T (Value Replacement Tool)
Purpose: Extend Gen-T’s table reconstruction process to replace unsupported or unreliable values in source tables.
​​​​​
Work:
Gen-T is a tool that retrieves tables from a large data lake that are related to a specified source table and reconstructs the source using various merging operations on external data. From this reconstructed output, I extracted a support score for each table cell that represents how confident the system is in the correctness of that value.
Rather than relying only on exact string matches and simple counts, I designed an extension that uses semantic similarity and row-level context to evaluate how closely a candidate value corresponds to a given source value.
​​​​​
​​​​
Result: Gen-T now reconstructs tables by selecting values with the highest semantic support rather than simple lexical matches. This allows the system to recover more accurate and complete tables even when values differ syntactically, are partially missing, or have been transformed, making it more robust to noisy and fragmented real-world data.
​
Link: Report