Machine Learning Engineer focused on getting models into production and keeping them reliable once they're there.
About Me
I'm a Machine Learning Engineer with a background in Computer Engineering, focused on building and deploying machine learning systems. My experience spans classical ML, deep learning, and NLP, with an emphasis on taking models from experimentation to production.
I've worked across public healthcare, automotive research, and international development, building data pipelines, APIs, and monitoring for real-world ML systems. I'm particularly interested in MLOps and production machine learning. Outside of work, I play chess, contribute to open-source projects, and enjoy long walks with my dog.
Work Experience
Production ML work across healthcare, automotive, and international development
Machine Learning Engineer
Oct 2024 – Jan 2026- ▸Automated the OCR workflow (OpenCV, Tesseract, PaddleOCR), cutting processing time from 20 days to 7 hours.
- ▸Built NLP pipelines (Kedro, DVC, MLflow) for automatic classification of medical reports, ensuring full audit traceability.
- ▸Shipped FastAPI services on Docker with MinIO as the data/artifact store, and standardized the model lifecycle with MLflow tracking and a remote registry.
Data Scientist
May 2023 – Sep 2024- ▸Built ML models (Scikit-learn, Pandas, NumPy) that contributed to an 18% increase in customer satisfaction.
- ▸Implemented NLP pipelines (NLTK, spaCy, fastText, Word2Vec) to automate insight extraction from text.
- ▸Integrated ML models into legacy C# systems and ran A/B tests to validate new solutions against the incumbent.
Researcher & Data Scientist
May 2023 – Apr 2024- ▸Analyzed EV performance data with PySpark and MLlib, applying hypothesis testing and multivariate analysis.
- ▸Built regression, classification, and clustering models on telemetry data in a GCP environment (BigQuery, Vertex AI).
- ▸Presented results to international leadership, collaborating with engineering and business teams on feature modeling.
Data Analytics Intern
May 2022 – Nov 2022- ▸Supported international data analysis projects with multicultural teams at a UN migration agency.
- ▸Built ETL pipelines (Pentaho, SQL, SQLAlchemy) and API/dashboard prototypes (FastAPI, Streamlit) for stakeholder reporting.
- ▸Managed data across Azure and S3, and contributed to Power BI dashboards for decision-making.
Skills & Technologies
Tools I actually use, grouped by where they show up in the work above
Languages & Frameworks
Python, SQL, C#, Pandas, NumPy, Scikit-learn, TensorFlow, PyTorch, FastAPI
MLOps & Cloud
Docker, MLflow, DVC, Kedro, Airflow, Spark, AWS, GCP, Azure
Databases & Visualization
PostgreSQL, MongoDB, Matplotlib, Seaborn, Plotly, Power BI
Featured Projects
Some of the projects I've developed
AI Hiring Matcher
Resume-to-job matcher that audits its own training label for bias before trusting it. It found gender skew up to 89 percentage points by role, and is deliberately designed not to reproduce it. Evaluated as closed-set retrieval across 51 known roles (88.2% Recall@5), not generalization to new postings. Fully local MLOps stack (DVC, MLflow, Evidently), no cloud account required.
CiteRAG
Self-hosted RAG system for technical docs. If the model can't back an answer with a citation, it refuses instead of guessing. Most RAG demos skip this fail-closed design. Hybrid retrieval, cross-encoder reranking, fully local inference; nothing leaves the machine.
OCR Benchmark Pipeline
Containerized pipeline benchmarking OCR engines (Tesseract, EasyOCR, PaddleOCR) against scanned documents, scoring real accuracy (CER/WER, field-extraction, IoU localization) against ground truth, not just timing. New engines plug in through one interface; ships with sample data and a fully offline mode, so it runs end-to-end with zero setup.
Ride Demand Forecasting
Ride-demand forecasting rebuilt honestly on public data instead of reusing a company's take-home dataset. Caught a naive train/test split silently inflating the eval error by 4x, then fixed two more leakage bugs shipping the notebook as a real FastAPI service.
Get in Touch
Open to full-time roles, freelance work, and collaborations