Nursultan – Golang, Kubernetes, Kafka

Multi-Agent Opinion Simulation Platform

• Built a multi-agent opinion simulation engine where LLM-powered agents model stakeholder discourse and competing interests across iterative deliberation rounds;
• Designed agent cognitive profiles incorporating motivated reasoning, loss aversion, the Granovetter threshold model, and commitment escalation to simulate realistic opinion dynamics;
• Implemented strategic agent behaviors including POST, AMPLIFY, COUNTER, COALITION, REFRAME, and SILENT, enabling agents to adapt their behavior throughout simulations;
• Built a deterministic LLM-independent routing layer enforcing strict information isolation across belief, public, and private communication channels;
• Developed simulation evaluation and convergence tracking using Shannon entropy, Gini coefficient, Overton-window drift, and coalition stability, with stop conditions including polarization lock and strategic equilibrium;
• Implemented asynchronous multi-agent orchestration with concurrent LLM execution and support for both Claude and OpenAI providers;
• Added mid-simulation event injection to test agent adaptation and simulation robustness under changing conditions.

Platform engineering to automate deployment and operational processes across several Kubernetes clusters. Work spanned building and maintaining a large fleet of operators, backing trigger services and APIs, and contributing to an in-house framework that standardized and accelerated operator development — with AI agents assisting both development and cluster monitoring.

• Built 4 Kubernetes operators and maintained 26 across several clusters, automating deployment and operational processes.
• Developed trigger services, microservices, and API endpoints to orchestrate platform workflows.
• Contributed to an in-house operator-scaffolding framework tailored to the platform, cutting operator development time from 6 weeks to 2 weeks.
• Introduced AI agents to assist operator development and monitor cluster health, reducing manual ops toil.
• Designed and implemented a scalable multi-channel notification service supporting in-app, email fan-out.
• Managed and developed 4 senior engineers: performance reviews, growth plans, and technical mentorship.
• Facilitated cross-team design sessions to align API contracts and architectural standards, improving system interoperability.

Design and delivery of core banking APIs and microservices for a high-throughput financial platform, with an emphasis on scalability, real-time event streaming, and regulatory compliance.

• Architected core banking APIs and microservices with worker pools and async schedulers, sustaining 10k RPS at 99.9% uptime;
• Introduced gRPC and Kafka to decouple services and enable real-time event streaming;
• Used Redis for caching, prefetching, and rate limiting under peak load;
• Optimized PostgreSQL queries and indexes, reducing p95 latency by 60% on critical endpoints;
• Partnered with product owners and QA to refine SLAs and ensure compliance with banking regulations.

Loan and BNPL offer aggregation platform enabling users to submit a single application, compare offers from eight banks, and select the most suitable financing terms.

• Built transaction-safe financial APIs handling millions in monthly transaction volume;
• Automated audit reporting workflows, saving 40+ hours of manual work per month;
• Reduced average API response times by 45% through query batching and transaction restructuring;
• Collaborated across engineering and business teams to strengthen compliance, security, and financial data handling.

Data visualization platform for a Thai energy company, providing interactive UI tools for visualizing and analyzing oil and gas fields.

• Built backend services fetching time-series data from OpenTSDB and OData endpoints, aggregating into BigQuery for OLAP analytics across 4 oil and gas fields;
• Parsed Excel workbooks submitted by field engineers to model and predict output metrics across gas and oil production sites;
• Developed reporting and visualization pipelines on aggregated field data;
• Applied prior field engineering domain expertise to accelerate delivery and client alignment.