Lee Penkman

I build and run applied AI products across coding agents, model gateways, speech tooling, local memory, media generation, 3D tools, forecasting, and model acceleration.

Technologies: Go, Python, CUDA, Triton, PyTorch, React, FastAPI, Postgres, CuteDSL, Zig, Rust, C++,, ONNX, local GPU serving.

• Codex Infinity, OpenPaths, CuteDSL, GPUBrain, DictatorFlow, Text Generator, Netwrck, Papers.app.nz, SimplexGen, Helix, and related product systems.
• Production details I care about: coding automation, AI scaling, observability, provider routing, CI repair loops, deployments, and performance.

I worked on Microsoft Clipchamp, a video creation and editing product, collaborating with research teams on AI features that shipped into real creative workflows.

Technologies: PyTorch, Hugging Face, ffmpeg, NLP, deep learning, video processing, production ML.

• AI video creation.
• AI video editing.
• Background removal.
• AI VFX and media-generation workflows.

I was the first ML engineer at Compose.ai, improving the writing assistant, the data process, and the model iteration loop.

Technologies: PyTorch, Hugging Face, React, NLP, Weights & Biases, AWS, spaCy.

• AI autocomplete.
• AI writing assistant features for rewriting and rephrasing text.
• Spell checking, dataset curation, and bias reduction.
• Virtual assistants, chatbots, model auto-training pipelines, and LLM inference acceleration.

I delivered image processing and machine learning capabilities for Canva photo editing, after earlier infrastructure and security engineering work.

Technologies: Terraform, PyTorch, AWS, Docker, Python, GLSL/OpenGL, CI/CD systems.

• Machine-learning background removal and object detection.
• Dataset collection and curation pipelines for training ML algorithms.
• Phishing detection, infrastructure security, penetration testing, and incident response/on-call.

I led full-stack product, testing, analytics, data engineering, and warehousing work for high-traffic news products.

Technologies: Terraform, Django, AWS, Docker, Node.js, React, Google Ad Manager, Tag Manager, Optimize, Google Cloud.

• Rolled out paywalls to high-traffic news sites.
• Scaled and developed high-traffic publishing systems.
• Built end-to-end analytics pipelines, reports, and video/ad optimization workflows.

I supported animation, motion editing, crowd simulation, mocap, and virtual artists with production tools, data systems, and reliability engineering.

Technologies: Python, Node.js, React, Ansible, Docker, ELK, Grafana, Redis, RabbitMQ, Postgres, MySQL, Maya, Nuke.

• Data engineering, syncing, queueing, and reliability engineering for warehousing projects.
• Production tracking, product-management tools, analytics, batch/realtime warehousing, and report generation.
• Image-processing pipelines, camera warping, rendering workflows, and artist workflow improvements.
• Versioning, tracking, publishing, and diffing for 3D artwork, cameras, meshes, textures, shaders, skeletal motion, and time-series data.
• Distributed 3D data pipelines, format conversion, and performance optimization.

I worked on data products for the .nz namespace: registry statistics, broadband mapping, topology visualization, and student mentoring.

Technologies: AWS, Python, Puppet, Ruby/Rails, D3, Hive, HBase, Hadoop, Scala/Spark, Postgres, Elasticsearch, Docker, Elastic Beanstalk.

• Registrar stats dashboard.
• BroadbandMap.nz.
• Internet topology visualization in D3.
• Mentoring student projects in Hive, HBase, and Hadoop.

I worked remotely in an agile team on government service discovery products, and introduced Jasmine-based behaviour/test-driven development.

Technologies: Python, Puppet, Django, Docker, Bash, Less, Backbone, Bootstrap, PostgreSQL.

• Human Services Network for NSW Government.
• Service Seeker, a map-style service discovery product.
• Testing discipline that made regressions visible and easier to control.

I worked on responsive product engineering, cloud systems, distributed workflows, and performance-sensitive marketplace features.

Technologies: PHP, Go, JavaScript/jQuery, AWS, MySQL, MongoDB, SQS.

• Design annotations for commenting on specific parts of an image.
• Campaign Monitor to Mailchimp migration with no downtime using SQS.
• Data migration to a new SSO service.

I worked on special projects covering bug fixing, developer productivity, test automation, data warehousing, and business intelligence.

Technologies: Pentaho BI, Java, ExtJS, MySQL, MSSQL, Jenkins, ETL workflows.

• Data warehousing and business intelligence solutions.
• Migration of distributed/service-oriented systems to Spring Remoting and service discovery.

I helped teach Java, Python, and general computing. I like teaching because it forces clear explanations of how computers actually work.

Technologies: Java, Python, practical computing, debugging, teaching labs.

• Supported students across practical programming, debugging, and core computing concepts.

I worked through the full software lifecycle: requirements, mockups, client/server modelling, schema design, development, test automation, QA, and a final presentation.

Technologies: Java, ExtJS, MySQL, Maven, SVN, Tomcat, Hibernate, Spring, Balsamiq, Jenkins.

• Continuous Automated Scheduling Environment, a rule-based scheduling system.

I was a web developer apprentice at a small startup importing goods and building ecommerce websites to sell them.

Learned: LAMP, HTML, JavaScript, CSS, AdWords, Analytics, AdSense, SEO/SEM, marketing.

• Designed, built, and marketed ecommerce websites from scratch.
• Developed a flight-time calculator.