Technical lead for FPGA-based IMU SPI Bridge integration into HIL, enabling first-ever execution of production firmware in E2E simulation. Drove company-wide adoption and assumed ownership after senior departure

-> Architected multi-IMU simulation-to-board communication and deterministic fault injection (C++/Python)

-> Drove fault injection strategy from needs analysis through design, validation, and company-wide rollout

Refactored core test infrastructure: 60% runtime reduction (~9 hrs/day saved) and 75% codebase reduction, improving modularity and maintainability

Delivered 17 verifiers and 80+ tests in 3 weeks during milestone-critical crunch time

Designed and implemented 22+ sensor fault injection mechanisms across radar, lidar, and IMU emulators, including packet- and time-based activation/reset for high-precision system validation

Own health, execution, and release gating for 450+ HIL tests, supporting bi-weekly release candidates across log-based and simulated environments

Launched first end-to-end regression suites for perception and firmware features (e.g. impact detection, ultra-short-range detection, localization, environmental conditions, obstruction handling, fail-operational behaviours)

Built unified tooling to raise arbitrary faults on-vehicle and in HIL, validating fail-operational and safety-critical system responses (C++, Python)

Led team of 4 to migrate and simplify 60k lines of legacy code for new test infrastructure

Defined and implemented cross-team validation processes improving requirements-to-feature communication, hardware integration into HIL, test suite bring-up, regression validation, and milestone approval workflows

Co-developed MCP Server for Polarion, enabling AI-driven requirement summarization (including images) and automated code-reference discovery; delivered productionization strategy

Led development of robust, large scale map versioning system across Embark infrastructure (C++, Python)

-> Balanced needs, workflows, and constraints from 7 teams, implementing changes to 5 repos at Embark over 8 months

-> Results: used as standard for 3000+ of simulations run daily, across 3 platforms (local machine, cloud, CI/CD), and enabled rapid testing of new maps for autonomy at scale

Owned and maintained high-fidelity controls simulator at Embark, after Senior Engineer left

-> Drove investigation for alternative simulators, ultimately designing migration to new controls simulator for improved maintainability, robustness, and long-term feature growth

Used C++ templates to refactor and expand internal rosbag reader libraries to unlock multibag-processing for multiple data sources

-> Enabled ground truth bags to be integrated into simulation for the first time, allowing perception team to collect previously unavailable MOT metrics

Pitched 8 new projects which increased maintainability of stack, improved regression testing, improve readability of code

Led needs analysis for new non-technical users of simulation

-> Identified biggest problems, sorted by problem area and proposed solutions for each problem with relative priorities

-> Gained stakeholder and management approval for all 15+ solutions, all prioritized for Q1 2023

Completed 4-month mentorship under a Senior Program Manager

Was specially selected for 5-person strike team lasting 2-months, responsible for driving engineering efforts for first major customer milestone (Truck Transfer Program)

-> Established team’s new workflows with 9 teams to prioritize rapid testing iteration

-> Provided on-call support for trucks during round-the-clock efforts to accelerate truck testing and implementing or delegating bugfixes

Led design and implementation of new software architecture for modelling complex iterative algorithms (Anytime Algorithms)

-> Enables deterministic offline simulation of cross-compilation production environments, navigating complex timing issues

-> Key features: modularity, statelessness, performance tracking and adaptability

-> Implementation: Developed in C, complete with CMocka unit tests

-> Validation: Quantified extents of existing of platform via diagnostics collected using new framework, and made recommendations for further development

-> Technologies: C, Bazel, CMocka, Jenkins, Bash + Python scripts

Self-taught new robotics software framework

Led and delivered new software architecture within 2 months

-> Led design process and implementation (ie. explored problem space, identified requirements, designed multiple iterations, implemented design with unit tests)

-> Earned approval from senior software engineers on design and results

Pitched concept for new cross-platform product feature

Updated and developed new documentation for smoother technical onboarding

Independently led design, software architecture, and implementation to introduce a Localization Validation Framework to autonomy stack

-> To be used as foundational localization simulation, testing, quantification, and state-saver for perception stack

-> Created ROSBag database management system, identifying and categorizing into 16 localization incident types across 9 location types; became preferred resource manager for autonomy projects thereafter

-> Key features: complete modularity via config. files + database, automation, consistency, and state-freezing

-> Designed and implemented components: database (above), I/O data processing + aggregation, process management system (controls spawning + graceful termination of software elements), and simulator system

-> Technologies: ROS, Python, C++, Rviz, Airtable

Contributed to modifications, discussions, analysis and testing for localization stack to improve robustness in field

Conducted on-site overnight testing representing localization team (both independently and in group)

Pitched initial project to introduce automation to data analysis pipeline, and developed into multi-purpose simulation, testing and analysis framework (Localization Validation Framework)

-> Ideated feature list, scoped project deliverables, managed timelines, and designed solution

Led testing (DoE), engineering and productionization of LiDAR level calibration method to 0.4° accuracy; robust to unit variation between LiDARs