Dennis Biber

Software and AI engineer with a decade across autonomous vehicles (Uber ATG), AR/VR research (Meta Reality Labs), and defense robotics (DARPA / Carnegie Robotics), now building custom on-premise AI, embedded, and data systems independently. Specializes in local-first AI agents and computer-vision systems that businesses own outright — no cloud dependency or subscription lock-in. Combines production machine-learning and data experience with hands-on embedded and full-stack development.

Independent Technical Consultant at Independent Technical Consultant (2022-01 – Present)

Custom on-premise AI, embedded, and data systems for small businesses — locally owned, no cloud dependency or subscription lock-in.

• Built an autonomous "game master" AI agent for live escape-room operation — a function-calling agent that tracked player progress via camera input and triggered in-room actions and contextual hints when players stalled; delivered as a working proof of concept to augment on-site staff.
• Designed a computer-vision system for an experiential-entertainment venue that identified high-interest moments during live customer activity and automatically generated personalized highlight reels offered back to customers.
• Specified and built on-premise AI compute infrastructure (multi-core Xeon, 48 GB GPU) to run these systems locally with no recurring cloud or subscription costs — a local-first architecture I continue to develop and extend.
• Engineered ROS2-based control systems and Arduino-driven hardware for interactive physical environments, including a UWB-radio navigation "compass" and a custom themed arcade game.
• Delivered Python data analysis and visualization for a 2022 election-cycle organization, translating COVID, economic, and other datasets into clear briefings for non-technical stakeholders.
• Built and deployed a custom React web application for a community organization; provides ongoing technical advisory to a financial-compliance consulting firm.

Software Engineer & Test / Safety Lead at Carnegie Robotics (2021-08 – 2022-04)

DARPA RACER program — autonomous off-road ground vehicles for GPS- and comms-denied environments.

• Served as test and safety lead and primary point of contact for partner competition teams (CMU, JPL / Caltech, University of Washington), defining test procedures and training all teams on vehicle operation.
• Led on-site data-collection operations at Fort Irwin, managing a cross-functional field team (hardware technician, hardware engineer, software engineer) through a multi-week competition deployment.
• Ran standardized daily vehicle testing with structured stakeholder reporting; contributed to field repair and platform readiness.
• Integrated an OCR verification component into a C++ computer-vision targeting system on a follow-on defense contract.

Data Capture Technician at Meta (Reality Labs) (2019-10 – 2020-10)

Codec Avatars — photorealistic-avatar research; owned data-capture software and large-scale GPU processing pipelines.

• Operated avatar-generation pipelines across thousands of GPUs, deliberately throttling concurrent usage to 1,000–2,000 GPUs to balance throughput against shared-cluster demand.
• Built CI/CD pipelines for data-capture systems (hundreds of synchronized cameras, microphones, and environmental sensors), adding version control and diagnostics that supported scaling to 20 capture sessions per week.
• Developed cron-scheduled overnight Python test suites with an automated failure-reporting system that emailed the owner of any mechanism failing a test, surfacing hardware issues before scheduled captures.
• Handled real-time troubleshooting during live capture sessions and QA on processing results.

Test Engineer at Uber Advanced Technologies Group (ATG) (2015-11 – 2019-10)

Autonomous-vehicle program; progressed from vehicle test operations into software, data, and ML systems development.

• Led development of a Slack chatbot that generated and executed vehicle test configurations from natural-language input using GPT-2, productizing an internal prototype into a team-maintained tool.
• Built a Python model (linear regression on motion-planning data from hundreds of human drivers) to quantify autonomous-vehicle behavior repeatability; the method was adopted into the production motion-planning stack for lane-change logic.
• Served as test lead for an end-to-end deep-learned driving system developed with NVIDIA, managing three engineers and defining SOPs for neural-network fine-tuning and validation; presented the system to the CEO.
• Built an automated reporting pipeline that generated and routed stakeholder reports on each vehicle data-log offload; contributed C++ and in-vehicle UI work to the software-integration team.
• Developed Python data-analysis and classification tooling to grade incoming LIDAR and GPU hardware from bench stress-test data, standardizing hardware acceptance criteria.
• Built operational programs from scratch — fleet safety-operator training, software-release acceptance testing, and multi-city launch operations (San Francisco, Phoenix) — defined their SOPs, and transitioned each to less-technical owners who sustained them independently.

Associate of Science in Electrical Engineering Technology – Pittsburgh Technical Institute

Additional coursework toward B.S. in Applied Management – Pittsburgh Technical College