About Me
Hello! I'm Anh Kent, a passionate Full Stack Developer with a strong foundation in computer science and a talent for building efficient, scalable applications. My journey in technology is driven by a commitment to continuous learning, innovation, and contributing to impactful projects. Having an Agile mindset, I thrive in collaborative environments and enjoy taking on new challenges that help me grow both personally and professionally.
/ Core Values
/ Education History
Bachelor of Computer Engineering
2020 – 2024Ho Chi Minh City University of Technology
Capstone Project
Real-time AI-Integrated Security Camera System - A Video Surveillance as a Service approach
Paper
Authored a first-author research paper on accelerating emerging computer vision models for Video Surveillance as a Service (VSaaS) using Edge TPU — written but not publicly available.
Experience
atWare VN
August 2023 – Present- Building an AI avatar chatbot — an interactive assistant with an avatar persona, backed by an LLM. Users can speak to the avatar and it responds accordingly with in-persona voice replies.
- Integrated the chatbot into an existing production web application, fitting its architecture and UX without disrupting current users.
- Integrating AI models to drive the avatar's domain-specific, in-persona responses.
- Set up error monitoring with Datadog to catch and diagnose production issues early.
- Built performance tests with Artillery and Playwright, plus end-to-end test suites with Playwright.
- Building an offline-first Progressive Web App that stays fully usable on unreliable networks through service worker caching and background sync.
- Designed and implemented a sync engine using a Last Write Wins conflict resolution strategy to reconcile concurrent offline and online updates across devices.
- Implemented queued writes that replay automatically on reconnect, with rate limiting and exponential backoff for fault-tolerant sync.
- Implemented a CQRS backend, separating the command and query paths so offline write replays and reads scale independently.
- Fine-tuned open-source Small Language Models using Japanese domain-specific data for specialized tasks, terminology, and response styles.
- Applied LoRA/QLoRA (Quantized/Low-Rank Adaptation) parameter-efficient fine-tuning to reduce GPU memory and training costs while preserving model quality.
- Implemented LLM guardrails for input validation, prompt filtering, sensitive-information protection, harmful-content detection, and structured output enforcement.
- Built end-to-end workflows for model training, evaluation, deployment, and integration with existing infrastructure.
- Delivered a working PoC based on a fine-tuned open-source model with integrated guardrails.
- Built an AI chatbot that streams responses in real time over Server-Sent Events (SSE).
- Built a custom AI Agent feature — agent logic implemented in-house, invoking Claude and GPT models through the AWS Bedrock SDK — enabling multi-step task handling beyond plain Q&A.
- Persisted conversation history in MongoDB so the chatbot keeps context across sessions.
- Created a scheduling feature so the assistant can plan and trigger tasks at set times.
- Built a dedicated file-encoding microservice within the system's microservices architecture.
- Built a web-based admin app for monitoring and controlling video recording data feeds.
- Automated video retrieval and processing with serverless pipelines — AWS Lambda functions triggered on EventBridge schedules.
- Used Amazon SQS for fault-tolerant transport, sustaining a throughput of nearly 50GB of video daily.
- Designed a scalable CI/CD pipeline on AWS CodeBuild, improving deployment reliability and cutting build times.
- Migrated 100+ workflows from Jenkins to GitHub Actions for faster iteration and easier maintenance.
- Built an app for managing protected-area data, combining interactive maps with charts for statistical analysis.
- Worked directly with stakeholders to align the product with project objectives and user requirements.
// NDA — details are kept intentionally high-level.
Projects
$
Placeholder — these projects are sample data, not real work yet.
Certificates
$
Get In Touch
$ I'm always open to new opportunities and collaborations.
Feel free to reach out!