Piotr Poznanski

Developed a Flutter-based end-user mobile application for a parking entrance management system. Led a remote development team and created the system architecture for a distributed IoT product for managing parking spaces. The architecture included system software and hardware modules, data models, and protocols. Built a software toolkit for generating an object-oriented abstraction layer for embedded RTOS programming. Developed data acquisition and processing software for the Puma SE business intelligence unit.

Evaluated agentic frameworks such as LangChain, LangGraph, CrewAI, and AutoGen, as well as methods for knowledge extraction and graph generation. Developed a knowledge extraction and storage system. The information extracted from scientific papers was processed using methods such as ontology knowledge graph, RAGs, and Agents. Prototyped AI agent tools for the identification of new research areas. Tools based on AI agents and teams of agents were generating new research ideas based on information provided by the knowledge system. Researched methods for robotic arm navigation, including policy generation, reinforcement learning (RL), and various machine learning models like vision-language models (VLMs), large action models (LAMs), and large language models (LLMs). Prototyped a virtual test environment and tested an LLM-based approach for policy and commands generation.

Designed highly scalable multi-agent discovery and communication protocol to power a network of millions of AI agents running on a blockchain infrastructure. Built a prototype of agents implementing the designed protocol. Conducted preliminary scalability tests by running a network of thousands of agents, performing min-max tasks on multidimensional manifolds. Conducted research on new methods of data representation, learning, and inference. The research included topological data analysis, compressed sensing, deep neural networks, LLMs, high dimensional manifolds analysis, and field theory methods.

Created development plans and roadmaps for product implementation, including work breakdown structures and effort estimations. Provided guidelines for design and engineering team synchronization. Drove and provided feedback for product requirements formulation. Created and drove reviews of lo-fi wireframes and use-case diagrams for the first version of the MVP. Built a full-stack MVP team of six senior developers. Created a hiring plan. Led and participated in technical panels. Created the platform, product, and integration scalable architecture. The platform was deployed on AWS, heavily relying on AWS-provided services (RDS, Amazon Cognito, and ECS). It integrates with game engines, payment providers, and blockchains. Led implementation technologies and framework decisions. Back-end services were implemented in TypeScript and NestJS. The front end was implemented in TypeScript and Next.js. Architected and implemented a prototype (TypeScript/Node.js/React) of RBAC/ACL authorization and authentication core service. The service was a foundation for platform-wide user account management, shared by game engines and platform products

Created the architecture of a solution for cryptocurrency payments for using public wifi hotspots. The system comprises router extensions, Ethereum smart contracts, and a mobile application. Prototyped an Ethereum smart contract in Solidity for payments in a prepaid packages model with refunds on unconsumed data. Created and prototyped software architecture for wifi routers. The solution is based on OpenWrt, using integrations with a captive portal and filtering and quota management.

Architected an MVP for a startup banking solution. The architecture consisted of a Kotlin and Spring Boot back end, a Flutter front end, RESTful APIs, integrations with third-party core banking, and KYC and AML solutions. Gathered and analyzed requirements while working closely with the client and product owner. Evaluated third-party solution providers and oversaw the integration process. Led an entirely distributed team of 3-7 developers and collaborated with the product owner to provide a work breakdown and estimates.

Created and taught courses on software engineering, programming paradigms, programming languages, and object-oriented technologies. Mentored students and supervised dozens of graduate theses at the bachelor and master of science levels. Authored and co-authored 30+ scientific publications and successful scientific grant proposals. Participated in scientific R&D projects, including software for particle tracking in high-energy physics experiments, tracking-based software for generating physical motion graphics, and flood simulation systems. Created the technical vision, co-authored a successful grant application, and directed the technical aspects of an interdisciplinary project for creating the first Polish system for quantitative studies on democratic roll-call political systems.

Contributed to and supported the scientific project roadmap by developing and managing the technical agenda. Managed a team of 5-7 software and hardware developers in the CREDO project. Created high-level architecture for data processing in the CREDO project. Architected and co-developed a system and mobile application for detecting cosmic particles using phone cameras.

Built and led Agile teams of 7-14 developers in developing software for base radios for Motorola ASTRO trunked systems while serving as a technical leader and scrum master. Created the architecture and developed software components in C++ for application and service layers of Motorola base radios. Developed new methods for capturing and analyzing technical requirements. Took part in technical planning and project management for two major releases of Motorola ASTRO Base Radio software with release budgets estimated at $10-$20 million.

Developed substantial parts of the configuration management system in the Quattor toolkit. The toolkit managed CERN computer center production clusters for 10+ years and is still used by some high-energy physics (HEP) institutes. Co-created the pan configuration language and implemented the first version of its interpreting compiler in C++, Yacc, and Bison. Created and conducted workshops on the pan language and cluster management. Led 1-3 developers in the configuration management task of the European DataGrid project with a budget of over 70 million EUR. Created and taught courses on the pan language and Quattor toolkit when taking part in international conferences and workshops.