Khanh Nguyen

Senior software engineer, specialize in building production ML systems from Helsinki.

Helsinki, FI · khanhng93@proton.me · LinkedIn · GitHub

01 · About

Some context.

I'm a software engineer based in Helsinki with eight years of experience building production systems, from machine learning to backend infrastructure. I'm interested in building scalable ML systems for real world applications and exploring new technologies.

Beyond ML, I'm interested in systems design and developer experience.

Machine Learning
Inference
Data
Backend
Frontend
Infra

02 · Experience

Selected roles.

Senior Software Engineer QuantrolOx
2022 — Present
Espoo, Finland
  • Built an LLM agent as the agentic layer over complex hardware and quantum experiment workflows, natural-language control of multi-step engineering processes.
  • Built production ML systems for automated device calibration, reaching 99.8% gate fidelity and cutting calibration from ~1 day of manual tuning to ~30 minutes.
  • Designed and shipped a customer-facing Python SDK and internal tooling that became the primary developer interface across two research teams.
Machine Learning Engineer Vainu
2018 — 2022
Helsinki, Finland
  • Trained and deployed production NLP systems serving hundreds of thousands of daily requests at ~95% accuracy across seven languages.
  • Optimized inference for low-latency, cost-bounded serving with dynamic batching, cutting memory 40% while sustaining throughput.
  • Owned ML infrastructure on AWS and Kubernetes with automated deployment via GitLab CI/CD; cut deployment time from 4 hours to 10 minutes.
  • Ran crowd annotation end to end: designed labeling tasks and curated ~30K labeled examples into model-ready training sets.
  • Trained and deployed the first Finnish abstract summarizer, achieving state-of-the-art performance.
Software Engineer / Data Scientist Swappie
2022
Helsinki, Finland
  • Built an end-to-end ML pipeline on GCP (TensorFlow, FastAPI) and shipped demand-forecasting models from prototype to production with product teams.
Fullstack Developer RND Works
2017 — 2018
Vantaa, Finland
  • Trained the first Finnish word embeddings and built a native Finnish chatbot with Google Cloud speech/vision integration.

03 · Selected projects

Things I’ve built.

Speech-to-Text Inference Service

Profiled and optimized ASR inference with ONNX, CTranslate2, and 8-bit quantization for low-latency, cost-bounded transcription. Built a Conformer-based pipeline handling regional dialect variation, from raw-audio preprocessing to deployed FastAPI service.

ConformerONNXCTranslate2FastAPI

Vietnamese Legal RAG Assistant

End-to-end retrieval pipeline over Vietnamese legal documents: ingestion, BERT embeddings, FAISS vector indexing, and a chat interface that returns grounded answers with citations.

PyTorchBERTFAISSFastAPI

Invoice Processing Pipeline

Multimodal document-understanding pipeline combining OCR and local LLMs for automated extraction across diverse PDF and image formats. Privacy-preserving, deployed on-premise.

TesseractLocal LLMsOCR