Josh Niemelä

argmin.dk

josh@argmin.dk joshniemela joshniemela +45 50 32 45 42 /in/joshniemela /in/joshniemela

I bridge the gap between technical and non-technical stakeholders to deliver projects that more effectively incorporate business domain knowledge while minimising technical debt. I tackle a wide range of problems, such as performance optimisation, data modelling, and software architecture design. I also have extensive experience working with containerised environments and Linux.

Experience

Lead Developer - argmin Consulting

2023 ⟶

Pingo Documents

09/2024 ⟶

  • Performed a performance audit of a problematic part of the website, and reduced the number of database queries in MS SQL by 92% and decreased load times by more than 15x.
  • Implemented integrations to eTl (Danish land registry), CVR (Danish Central Business Register) to automatically fetch and display information needed by users.
  • Designed and implemented a major refactor to fix fundamental limitations in the existing codebase, making high-impact features possible to implement.

Two Scenarios

12/2024 - 12/2024

  • Built an integration for Two Scenarios to continuously transfer pharmaceutical pricing data into a database within a week, enabling them to meet a crucial deadline. See case study

AI Estate

06/2023 - 09/2024

  • Worked as a contractor for Pingo Documents (working with Pingo directly after 09/2024) and was responsible for completing the more complex tasks, as well as guiding and assisting two other developers in smaller tasks.
  • Responsible for CI/CD, DevOps, security and ensuring the team met deadlines for client deliverables.
  • Implemented major updates to UX and features, by making the frontend more responsive and giving it a more modern look as well as improving the previously unintuitive user flow.
  • Made a service to perform information retrieval on real-estate documents using OCR and a language model.

Promilist

01/2023 - 03/2023

  • Migrated a Python codebase to Julia, improving inference speed by 5x. This was used to calculate fuel-optimal paths for naval vessels, by graph-traversing space-time graphs on a geodesic map.
  • Implemented an API in Genie.jl to interface with the existing backend.

Machine Learning Developer - Juristic ApS

08/2022 - 02/2023

  • Developed a program to automatically digitalise handwritten relational charts, using a combination of computer vision, clustering and morphological image processing in two months.

KU Courses - Project

05/2023 ⟶

  • Developed a event-driven microservice based SPA in Typescript, Clojure and Rust which asynchronously scrape pages, stores and displays course information from KU's course catalogue with significant performance (3x-15x shorter latency) and UX improvements over the existing official solution.
  • PostgreSQL was used to store vector embeddings and course information.
  • ONNX was used to perform vector-based search of courses with a pre-trained sentence transformer model. This approach yielded more robust and relevant search results compared to the existing solution.
  • Used SEO and grassroots marketing to increase the monthly active users from 0 to 200.

Education

B.Sc in Machine Learning and Data Science

08/2022 - 19/06/2025

Copenhagen University (DIKU)
  • 7-point scale average of 10.9 / 12.0.
  • Took 30 ECTS of M.Sc courses in randomised and approximation algorithms, computational geometry, probabilistic machine learning and IT project management.
  • Wrote my bachelor thesis on topological deep learning and graph representation learning, which then evolved into a reproducibility study of state of the art graph machine learning models. It can be found on GitHub, BottlenecksWithinGNNs.

Skills

  • Languages: C# (.NET), Svelte/TypeScript, Clojure (JVM), Python, Rust, Julia
  • DevOps and Tools: Docker, TeamCity, GitHub Actions, VPS and cloud hosting (Hetzner, GCP)