Hello, I'm

Kaspian

Builder-Minded Engineer. Shaping AI, Web3, and automation systems.

Focused on practical implementation across agent workflows, productized infrastructure, and real deployment, with an emphasis on structure, usability, and long-term maintainability.

View Projects Download Resume

Open to engineering roles, research collaboration, and product builds.

Featured Projects

Selected systems with clear problem, implementation, and outcome.

Representative builds across AI knowledge workflows, workflow automation, productized Web3 systems, and autonomous on-chain execution.

Productionized

AI Knowledge Base

A retrieval system for private documentation, research notes, and multi-source reasoning.

Built a RAG system with multi-knowledge-base routing, citation tracing, and observability, focused on solving accuracy and auditability issues in real financial-system usage.

Designed chunking, hybrid retrieval, and reranking for higher answer precision on domain-specific corpora.
Added answer citations, source tracing, and prompt controls to make outputs more auditable for serious use.
Shaped the system for extensibility across API, ingestion pipelines, and front-end chat workflows.

Hybrid retrieval + citations + multi-tenant knowledge routing

LLM (Minimax API)PythonFastAPIPostgrespgvector

GitHub Details

Open Source

BizCover Automation Engine

A commercialization-oriented automation engine for insurance quote workflows, data mapping, and recoverable execution.

Built around the BizCover Australia quote flow, this system turns client intake, field normalization, and multi-step browser execution into reusable automation infrastructure rather than a one-off script.

Normalized Excel client records into a stable field contract, using mapping layers to absorb business variance and reduce maintenance overhead.
Split occupation selection, customer details, and business information into explicit workflow steps with clearer state boundaries for observability, retries, and recovery.
Scoped the next layer around quote extraction, execution tracing, and decision analysis so the workflow can evolve toward agent-driven execution and closed-loop optimization.

Automation infrastructure for turning repetitive quote operations into measurable, recoverable execution

Node.jsPlaywrightWorkflow AutomationData MappingOperations

GitHub Details

Live

ChronoStamp

An on-chain credential protocol and productized Web3 build for real-world event scenarios.

Built around real events and identity accumulation, combining off-chain eligibility checks, on-chain credential issuance, and user interaction into one coordinated Web3 product.

Used off-chain signature authorization with on-chain contract verification to create a clear claim flow with stronger execution reliability and verifiability.
Abstracted credential issuance through a factory pattern, enabling standardized deployment and extension across multiple event scenarios.
Integrated IPFS metadata storage, front-end interactions, and on-chain execution feedback into a complete product loop.
Designed for composability, portability, and future evolution, leaving room for later Agent + Wallet automation workflows.

A Web3 credential system designed for real product delivery and on-chain execution scenarios

HardhatNext.jsArbitrumIPFSEthers.js

GitHub Details

Open Source

Web3 Agent

An execution agent for on-chain tasks, wallet actions, and event-driven automation.

Built an agent around on-chain event monitoring, tool calling, and task execution, with emphasis on execution safety, state management, and full-process automation.

Structured tools and state so the agent can safely compose wallet actions with deterministic guardrails.
Integrated event triggers and execution logs for better visibility into multi-step on-chain workflows.
Made engineering tradeoffs around signing, retries, and environment isolation to balance speed, performance, and real-world feasibility.

A tool-calling on-chain copy trading agent system with execution boundary controls

TypeScriptPolygonAgentRPCAutomation

GitHub Details

Experience

Engineering work shaped by systems thinking and implementation rigor.

A mix of product, research, and hands-on build experience focused on intelligent systems and automation.

2021 - 2023

Backend Engineer / FinTech Systems Developer

China Merchants Bank / Information Technology Department · Chengdu

Worked in a real banking production environment on fintech systems, delivering business automation, data analysis, and intelligent risk-control workflows from requirement alignment through release and post-launch operation.

Contributed deeply to the Hailuo RPA+AI platform, developing and shipping backend services on the bank’s cloud-native platform to support hundreds of live automation scenarios in real operational workflows.
Integrated OCR, OpenAPI-based capabilities, low-code patterns, and later-stage model interfaces so business requirements could be turned into usable, manageable internal systems with faster delivery cycles.
Built data pipelines and service layers around Kafka ingestion, Elasticsearch analytics, and reporting APIs, powering operational dashboards plus automated daily, weekly, and yearly reporting for product, business, and management teams.
Independently delivered an intelligent post-loan transaction-screening workflow, handling business alignment, project setup, development, release, and iteration while using XXL-JOB for large-scale concurrent processing under compliance constraints.
Built practical engineering discipline across project management and DevOps-style delivery, including requirement breakdown, multi-environment coordination, testing, release, and production issue handling, which shaped my long-term interest in transaction systems, capital flows, and automated execution.

JavaSpring BootMyBatisKafkaElasticsearchRedisXXL-JOBDevOps

About

A perfectionist drawn to restrained minimal interfaces and clear, practical engineering systems.

I mainly work at the intersection of AI systems, quant tooling, and developer products, with a long-term focus on how finance, trading systems, and automated execution can be engineered under real constraints.

My background in bank fintech delivery and financial systems makes me care about how technology becomes something runnable, scalable, and maintainable. The 2024-2025 period at UNSW in Australia then became the point where that engineering background shifted toward AI-native products, Web3 systems, and independent building.

Building deployable LLM applications around retrieval, evaluation, workflow orchestration, and automation.

Following frontier AI products, emerging concepts, and category-defining companies closely enough to turn daily learning into product and engineering judgment.

Exploring Web3, DeFi, prediction markets, wallet-connected products, and automated execution systems.

Using independent building as a long-term path from traditional finance into AI- and crypto-native software.

Currently focused on agent reliability, execution-state design, and data-driven automatic feedback loops. I want to bring experience from traditional financial systems, risk control, and capital-flow processing into AI software, distributed finance, and automated trading, and I remain especially interested in how Stripe-like product discipline and the Circle stablecoin narrative may shape the future of financial software.