Overview

Machine learning engineer focused on deep learning and agentic AI, with hands-on experience taking systems from research prototypes into deployed products. Work spans tool-using LLM workflows, cloud-delivered AI services, and healthcare NLP research, backed by a software engineering foundation in performance-sensitive simulation systems with a strong focus on reliability and maintainable design.

Experience

Selected work across agentic AI, research, simulation, and graph systems.

Microsoft

Lead ML Engineer (Secondee Via UST) · Applied AI systems

  • Designing and delivering production-grade AI workflows from early concept through deployment.
  • Building agent-driven systems with strong emphasis on reliability, traceability, and human oversight.
  • Collaborating across engineering teams to improve platform quality, observability, and secure system design.
Agentic AICloud AI ServicesSystem DesignObservabilitySecurity Engineering
📥Input🧠Planner📡Fetch DataAnalyzeReviewer📤Output
Idle

Thesis Research

MSc AI Research · Radiology report summarization

  • Developed a two-stage fine-tuning pipeline for clinical report summarization using LLaMA and T5 variants.
  • Built preprocessing and supervision pipelines for longitudinal radiology narratives with strict signal preservation goals.
  • Expanded evaluation beyond summary overlap with BLEU-4, ROUGE-L, METEOR, BERTScore, and RadGraph-oriented quality checks.
PyTorchHugging FaceLLaMAT5Clinical NLP
Radiology Report
CT chest without contrast demonstrates no focal consolidation, pleural effusion, or pneumothorax. The heart is normal in size. Mediastinal and hilar contours are unremarkable. Several small sub-centimeter lymph nodes are noted in the mediastinum, likely reactive. No suspicious osseous lesion identified. Limited evaluation of the upper abdomen is unremarkable.
LLM
LLaMAT5
Summary
Waiting...

CAE

Immersive Environment Developer · Simulation systems

  • Diagnosed and resolved high-impact defects in a real-time flight simulation codebase to improve stability and root-cause speed.
  • Implemented avionics and environment features under strict real-time constraints, including LiDAR-style scene encoding.
  • Developed automated regression tests, packet-level diagnostics, and performance optimizations for internal tooling pipelines.
C++C#SpecFlowWiresharkPython
6×6 world · 8 objects

Conova AI

Full-Stack Intern · Graph and localization tooling

  • Built an internal Neptune administration console with end-to-end graph entity CRUD workflows.
  • Implemented serverless API integrations for graph operations and user-group management.
  • Automated translation and localization ingestion by integrating external API services into backend workflows.
AWS LambdaAmazon NeptuneREST APIGraphQLTypeScript
WORKS_ATWORKS_ATKNOWSKNOWSKNOWSAlice:PersonBob:PersonAcme:OrgReact:SkillPython:Skill
API Log
Waiting for requests...

Personal Projects

Project work spanning healthcare NLP, adaptive streaming, computer vision, and reinforcement learning.

Technical Skills

Core technical areas reflected from current engineering and research work.

Programming Languages: Python, C++, C#, TypeScript, JavaAgentic LLM Systems: Microsoft Agent Framework, Azure AI Foundry, GraphRAG, grounded retrievalMachine Learning: PyTorch, Hugging Face, Azure ML, model tuning and evaluationAPIs and Data: REST, GraphQL, SQL, graph and NoSQL data platformsCloud Delivery: Azure and AWS deployment, monitoring, and automation