Founding technical leader of an agricultural robotics company building an autonomous mechanical weeding robot for organic farms. Owner of every piece of software on and around the vehicle - embedded control, perception, navigation, simulation, telemetry, cloud, and developer tooling - along with the sensor and actuator packages and the technical direction of the engineering team. Started in a garage with the CEO, through YCombinator (W25), to $2.7M raised.

• Lead the software engineering function: technical direction, hiring, on-boarding, and performance management, and delivery of the web operator interface and autonomous navigation controller.
• Architect and principal author of the robot codebase - about three quarters of all commits - a Rust system spanning services, motion control, sensor and actuator drivers, telemetry, and simulation.
• Own the sensor and actuator packages end to end: requirements, vendor sourcing, hardware selection, and integration, through to GMSL camera bring-up, optics, illumination, and shutter-synchronized strobing.
• Built the perception post-processing stage that turns noisy per-frame detections into stable tracks, suppressing identity flicker and box-size variance for downstream control.
• Wrote the controller for a high-inertia rotary weeding tool, solving motor timing and synchronization while both the target and the terrain underneath it keep changing.
• Developed the tool-to-camera calibration and the four-bar linkage kinematics placing the implement in height and reach - error-prone given a tool deliberately kept invisible to perception.
• Built the company's engineering infrastructure from nothing: CI/CD, release tooling, Terraform-managed cloud, WebRTC teleoperation relay, and TURN servers, and the corporate network.
• Set a systems-integration-first hardware strategy, buying hardware wherever the market had an answer so engineering effort concentrated on the software that differentiates the product.
• Regularly in the field with growers, turning what they actually do into technical requirements rather than our own assumptions about their needs.
• Took the platform from garage prototype to the second-generation vehicle running today; now driving a redesign targeting 1 m/s with greater than 90% weed clearance between crop rows.

Part of the Integrated Software group, responsible for all parts of the company's rocket software stack. Focused on the next generation of flight software, and specifically on making it testable, verifiable, and deployable at every stage - from a developer's laptop, through simulation and hardware-in-the-loop rigs, to an integrated vehicle. Treated a launch as the last instance of a procedure already run hundreds of times in test rather than as a singular event, and pulled practices, tooling, and languages in from outside aerospace so that critical software could change safely and often.

• Evangelist for new processes, tools, languages from outside the aerospace sector to improve team velocity and reduce friction for cross-team collaboration. Advocated for standardizing across the software team to help team member fungibility, ease on-boarding of new members, and expand the hiring pool of new candidates.
• Developer of the standard procedures and run-book for test and flight operations, making a launch a special case of procedures already run many times in test environments.
• Identified and remediated performance bottlenecks in the flight computer runtime configuration across the software implementations and the platform configuration.
• Contributed to building simulation proxy components to allow Vehicle-in-the-Loop testing, to help validate and ground the Hardware-in-the-loop tests on an integrated vehicle.
• Built the state machine validation engine for the in-house domain-specific language used for automation.

Responsible for all on-vehicle mapping contents for autonomous vehicles - from acquisition, to distribution, and back to telemetering discrepancies between perception and expectations.

• Architect of the in-car map distribution topology and pipeline: a data syndication model for location updates around the Ego vehicle, and a request-response model for route queries, serving very different radii and data types.
• Synchronized a real-time critical, highly scheduled piece of code with an asynchronous downloader and file manipulator process, including writing all libraries for downloading, decompressing, filesystem storage and retrieval, and content verification/validation.
• Member of the small committee defining the NVIDIA mapping schema across all layers; author of the grammar engine (Bizarro) and its semantic validators, holding the schema consistent across terabytes of map data.
• Built a continuous integration and deployment system on Gitlab's CI/CD to ensure code quality in Bizarro and across maps.
• Mentored other members of the team with in-depth code reviews, 1-on-1 coaching, and broader philosophical discussions on how to develop and maintain highly critical, human-safety relevant software.

Responsible for all aspects of Telemetry, Tracking, and Control ground software systems and architecture. As member of the team, reviewed code and architecture direction from peers. Principally responsible for hardware/vendor abstraction antenna dish controlling software to integrate with company standard software stack, and the antenna time scheduler for allocating communications with satellites. Built constellation network simulation software from scratch using data collated from public sources for analysis to optimize business operations and deployment plans.

• Built a Geo Information System database and analysis system for siting the gateways bridging the satellite network to the terrestrial internet, distributing traffic so no single ground connection saturated. Became a critical data source across satellite, network, and business operations.
• Designed and deployed the satellite communication task scheduler prioritizing and dispatching traffic between satellites and ground data centers - health monitoring, telemetry and ephemeris updates, and software updates in flight.
• Built a general-purpose antenna dish operating application extensible to any vendor's hardware over any interface, with proprietary algorithms for reliably acquiring satellites despite unpredictable sources of error.
• Directly supported satellite operations with software by providing many mission critical tools to ensure the safe and reliable operation of vehicles in space.
• Overhauled MVP architecture used for demonstration mission into a production ready global satellite internet service provider back-end capable of servicing billions of users.
• Architected a complete software defined network interoperating with TCP/UDP alongside a proprietary space-optimized protocol built for variable-delay satellite links.
• Lead investigator across control and data plane traffic, finding where and why radio throughput suffered, working with the Satellite Broadband Communication Integration team on ground-side and bidirectional aiming challenges.
• Co-authored the self-alignment algorithm for in-house Ka radio dishes and drove it from inception to a product deployable worldwide on power and network alone - system design, firmware, hardware and data plane control, and deployment debugging.

Built and managed two inter-related software teams delivering the firmware and algorithms for the Vessyl smart cup; owned the technical specification, selected major hardware, and mentored the junior developers.

• Designed and built from scratch a command-line IDE that combined Visual Studio, MSBuild, Keil µVision, MDK-ARM, and Git for cohesion between developers and workstation configurations.
• Crafted a chip vendor independent, event driven, embedded operating system focusing on power efficiency, featuring a decentralized cross-library signalling mechanism and extensive debugging capabilities.
• Developed core IP signal processing and machine learning algorithms for drink identification and nutrient detection.
• Wrote a MatLab based visualization front-end for demonstrating the technology to investors and press, and for internal testing of sensor hardware iterations.

• Developed a low-level fault-tolerant mesh network protocol based on IEEE 802.15.4 for use in and around automobiles.
• Future technology tolerant architecture, based around a unified message passing protocol.
• Designed and implemented many diverse sensors to capture valuable operating parameters of a vehicle.
• Tested the system by mounting it to a go kart and used for telemetry and performance tuning.

• Designed and developed a robust energy metering device for replacement of conventional electrical sockets.
• Incorporated an ad-hoc mesh network of inexpensive, low power nodes to communicate information to a central server.
• Custom PCB development, with focus on high signal integrity and ultra-low power usage.
• Performed mechanical and electrical design.
• First place winner of 2010 IBM Smarter Planet award, and Second place of 2010 Lockheed Martin Eta Kappa Nu award.

• Lead a small team to design, build, and test a technology designed to improve safety of cars while simultaneously improve the efficiency of driving habits.
• Designed a full sensing and processing package to be fitted onto a small 1/10th scale car design.
• Presented project at the Carnegie Mellon Undergraduate Research Symposium.