Technology leader for AI products and platforms.Hands-on architect, product planner, and engineering leader.
Summary
I am a hands-on architect who designs AI products and platforms and implements their core systems. Starting with analytical engines and server development, I went on to lead the launch and operation of B2C mobile services and B2B API services for institutions, as well as an R&D organization. Throughout this work, I have translated business requirements into data models and system architectures.
For a financial analysis AI system, I separated the model’s interpretation of questions and response generation from deterministic server-side computation, validation, and execution controls. I incorporated model evaluations and institutional proof-of-concept feedback into the design. In a subsequent project, I integrated multiple AI capabilities using a domain ontology and shared context. The company deployed the main capabilities to a testbed for review and testing by financial institutions.
AI · LLM
LLM Application Development
NLU
Tool Calling
MCP
RAG
Context Engineering
Model Evaluation
Fine-tuning Experiments
Architecture · Development
Software Architecture
Backend Development
C++
CUDA
Rust
Java
Python
Data · Cloud
Data Modeling
Vector Database
SQL / NoSQL
Azure
Google Cloud
Product · Leadership
Product Planning
Engineering Leadership
Qualification Highlights
LLM application architecture and core implementation. I personally carried out most of the core design and implementation of a financial analysis AI system. After identifying limitations in fine-tuning, I introduced retrieval-based intent analysis, then shifted to tool calling as model capabilities improved.
Agent execution with authorization and validation. I separated planning from final generation and implemented checks for permissions, provenance, and human approval requirements before execution. I validated structured outputs, failure fallbacks, and execution tracing in an internal integration environment.
RAG and shared data context. I implemented embeddings, vector search, and ingestion of source material for retrieval. In a subsequent project, I used a standardized data model and shared context to integrate and validate customer consultations, staff reports, consultation materials, and advisor support.
Model evaluation, error analysis, and regression improvement. I automated checks of context generation in multi-turn conversations against expected results. Domain experts led a manual review of all question-and-answer results. Comparisons of commercial and open-source models and tool-calling regression tests informed further improvements.
Platform development from engines to cloud deployment. I implemented analytical engines, asynchronous servers, streaming responses, and storage integrations. I used CUDA for parallel computation of annual cash flows. In a subsequent project, I validated GCP deployment and recovery procedures in an internal environment.
Institutional technical validation, regulation, and security. I incorporated financial MyData infrastructure and information security requirements into Azure infrastructure design and implementation. I led the technical delivery of institutional proofs of concept and reflected regulatory and security requirements in deployment architecture, providing technical support for adoption reviews and assessments.
Commercial service launches and client implementation. I was responsible for launching mobile pension advisory and institutional API services, including Azure deployment and operation. I helped convert an institutional solution proposal into a contract and led its integration into the client’s app.
Technical decision-making and hands-on engineering leadership. I led an R&D organization while continuing core development. I managed development teams, hiring, and performance reviews, and made the final technical decisions on architecture and deployment.
Experience
Approximately 16 years of formal employment
Head of R&D, CTO
Approx. 6 years
Joined as a team lead; subsequently led R&D as Head of R&D and CTO
I led the R&D organization and its technical decision-making, with responsibility for product planning, architecture, core development, and technical operations across financial analysis AI, mobile services, and institutional services.
AI product design and improvement: I set feature priorities and architecture for the financial analysis AI system. I directly handled data preprocessing, labeling, model training, and evaluation, and developed vector database-based intent analysis and tool-calling architecture. I used evaluation findings to revise prompts, tool descriptions, and response structures, then retested the changes.
Institutional proofs of concept, regulation, and security: I oversaw institutional requirements and technical delivery, incorporating evaluation feedback into the product. I documented environment separation, private connectivity, and access controls as deployment principles and implementation guidance, and handled technical work for innovative financial service designation reviews.
Cloud architecture and implementation for regulatory requirements: I was responsible for designing and building Azure infrastructure and related systems to meet financial MyData infrastructure and information security requirements. I supported infrastructure assessments and led the technical completion of external standard API connectivity tests.
Computation engines and service launches: I directly developed Rust-based cash flow and personal financial management engines and implemented portfolio and product recommendation rules defined with domain experts. I was responsible for launching and operating mobile analysis, advisory, and transaction capabilities, from account opening to closure.
Client implementation and operational recovery: I organized internal and outsourced teams, coordinated client requirements, and led the integration of an institutional solution into a client app. When computation errors occurred in the Azure-based service, I restored normal operation by fixing the core library myself.
Engineering management: I improved collaboration and shared understanding through cross-team code reviews. I evaluated new AI technologies, applied them to selected features, and shared the results through team training and development guides.
Software Development and Team Leadership
Approx. 10 years
Joined at manager level and led the development team
I planned personal financial planning solutions and developed their computation engines and services, establishing the foundations in data and business modeling that I later applied to AI products.
Analytical engines and business models: I designed and developed a C++ cash flow forecasting engine that connects income, expenses, assets, liabilities, and financial goals. I incorporated scenario comparison with adjustable assumptions into the financial planning solution.
Service development and operation: I worked across Java and Spring-based interfaces and servers, databases, batch processing, external financial product integrations, deployment, and operations. I contributed to the design and implementation of an online asset planning system and the application process for a subsequently granted patent.
Java-based Systems Integration
2+ years
Freelance
I worked as a freelancer on enterprise systems integration projects, developing Java-based integrations. Freelance work is excluded from formal employment tenure.
Patents
Granted patent related to online asset planning. As a co-inventor of a company-owned patent, I contributed to the system’s overall design and implementation and to the patent application process.
Technical Education and Mentorship
Technical training and AI adoption: I shared architecture proposals and development guides, and trained the team on AI technologies I had evaluated, model-specific use of Copilot, and reviewing and debugging generated code. This supported the team’s experiments with AI-assisted development.
Code reviews and collaboration: Cross-team code reviews improved mutual understanding of responsibilities and implementation scope. For some tasks, teams could identify the required implementation work and proceed without waiting for confirmation from the other team.
