About

I build production AI systems — LLM platforms, AI Governance, inference infrastructure, and the Kubernetes plumbing underneath.

Portrait of Si Thu Phyo

Senior AI Engineer with 7+ years shipping production AI systems — OCR, RAG, computer vision, and now LLM platforms and AI governance.

My work sits at the infrastructure layer: LLM gateways, agent runtimes, evaluation harnesses, observability, and the Kubernetes plumbing underneath. I care about the engineering problems that show up after the model works — routing, deployment, governance, cost, reliability.

At Arcfusion.ai I am leading ArcFoundry · delivering enterprise AI governance infrastructure for financial services. Alongside the platform work, I am exploring domain-specific language models and RL post-training (reward modeling, preference tuning, DPO/GRPO-style techniques) — the direction I want to grow into next.

Before Arcfusion: ML engineering at Carro (OCR, RAG, damage detection) and platform work for the UN World Food Programme via DIGITHAI.

Experience

  1. Senior AI Engineer at Arcfusion.ai
    Jan 2025 — Present
    • Leading ArcFoundry · delivering enterprise AI governance infrastructure for financial services.
    • Built a Chat-with-Data platform that lets users ask questions in plain English across several different business databases.
    • Set up the evaluation pipeline that runs before every release — catches regressions, ambiguous questions, and bad SQL before they reach users.
    • Ran the model-serving infrastructure on Kubernetes — GPU autoscaling and scheduling that keep latency and cost predictable at scale.
    • Built a meeting-intelligence platform with real-time processing that handles many concurrent sessions.
    • Wrote the deployment templates, evaluation setup, and observability dashboards the rest of the AI team reuses across projects.
    LangGraphAI GovernanceKubernetesFastAPIPGVectorLangFuse
  2. SDE & AI at DIGITHAI (UN World Food Programme Project)
    Jan 2024 — Dec 2024
    • Contributed to a large-scale Django platform supporting global food distribution workflows.
    • Designed secure card-status tracking and PIN management services with strict data integrity and audit logging.
    • Collaborated on system architecture and performed structured code reviews to improve maintainability and reliability.
    DjangoPythonSecurityAudit Logging
  3. Machine Learning Software Engineer at Carro
    Feb 2022 — Dec 2023
    • Improved CNN-based car damage detection model — reduced false negative rate by 50%, cutting manual inspection workload and improving pricing accuracy.
    • Built RAG-based internal knowledge assistant using vector search and LLM agents.
    • Designed OCR recognition pipeline (CRNN + Attention + ViT) trained on ~300,000 samples; achieved Levenshtein distance 0.084 on structured ID cards.
    • Deployed OCR inference service on AWS EKS with autoscaling pods and REST APIs.
    PyTorchOCRRAGAWS EKSComputer Vision
  4. Co-Founder & Lead Engineer at Rhythm18
    Feb 2019 — Present
    • Built and scaled a music education platform serving 10,000+ active students (Android/iOS + Django REST backend).
    • Led product roadmap and backend architecture from idea → launch → regional adoption.
    • Mentored engineers and delivered scalable APIs supporting user growth.
    DjangoMobileProductSelf-employed

Skills

AI Platform Engineering

AI Governance · LLM Gateway · Agent Runtime · Evaluation Frameworks · Observability · Cost / Quota Controls

LLM Application Engineering

RAG · Chat-with-Data · LangGraph · LangChain · PGVector · SQL Validation · Prompt Engineering

Production Engineering

FastAPI · Go · PostgreSQL · Kubernetes · Docker · AWS · CI / CD · OpenTelemetry

Education

Certifications