You're viewing the new portfolio — View classic site ↗
Atishaya Maharjan

Works

Professional experience, research contributions, personal projects, and technical skills.


Professional Experience

Mar 2024 – Present

Senior Software Engineer

LG Electronics · Englewood Cliffs, USA (Remote)

  • Maintained and built new features for B2B microservices serving 1,000+ vendors using Next.js, Redux, and TypeScript, driving a 25% increase in user engagement.
  • Implemented analytics and custom tagging frameworks to capture detailed performance metrics, enabling data-driven decisions across 6+ product teams.
  • Developed an AI search engine for the global GNB header, improving search efficiency by 40% and reducing latency by 35ms on average.
  • Engineered promo code functionality with magic link integration, increasing conversion rates by 15%.
Jun 2025 – Jan 2026

Senior Software Engineer

Hoek Insights · Amsterdam, Netherlands (Remote)

  • Built a Python package to embed interactive Three.js visualisations in Jupyter Notebooks via React renderers, cutting render time by 38% and reducing bundle size by 25%.
  • Designed an MVP marketing platform using FastAPI and PostgreSQL with 18+ endpoints, 10 workflows, and 95% integration-test coverage; achieved <200ms median API latency in staging.
Aug 2019 – Apr 2024

Software Engineer

Timeero · Phoenix, USA (Remote)

  • Led frontend development of mileage-tracking application using TypeScript, React.js, Vite, and Mapbox, serving 1,000+ active users.
  • Implemented 15+ geo-spatial features for segment mapping, improving time-capture accuracy by 30%.

Research

Mar 2024 – Present

Researcher — Algorithms & Graph Theory

GADA Lab · University of Manitoba · Supervisor: Dr. Stephane Durocher

  • Delivered talks at weekly GADA lab meetings on wiggly permutations (CCCG 2025) and space-efficient Eulerian circuits (SOSA 2026).
  • Developed a fast approximation algorithm for Hamiltonian path optimisation in spatially embedded graphs; built visualisation and randomised test-generation tooling across 100+ synthetic instances.
  • Contributed to noncrossing spanning trees library, improving performance by 30% and establishing Python best practices.
  • Engineered hypergraph visualisation tool for ML research.
May 2025 – Present

Machine Learning Researcher

Terrabyte ML Research Group · University of Manitoba

  • Extended Graph Neural Diffusion architecture to Hypergraph Neural Diffusion networks using PyTorch and PyTorch Geometric, prototyping 4 model variants.
  • Implemented MLFlow metrics and tracked 25+ experiments across 5+ datasets for reproducibility.
Jan 2024 – Jul 2025

Researcher — Combinatorics & Graph Theory

Department of Mathematics · University of Manitoba · Supervisors: Dr. Mahsa Nasrollahi-Shirazi & Dr. Andrii Arman

  • Proved a closed-form solution for the independence number of perfect binary trees and verified Hurlbert–Kamat's Conjecture using generating functions and computational analysis across 1,000+ cases.

Projects & Achievements

Brainstorm W

2026

Next.js · FastAPI · PostgreSQL · Pinecone · Redis · RAG

AI-powered, role-specific knowledge management system (144 commits, 4 collaborators). Integrated Pinecone, Jira, Redis, and PostgreSQL with interactive charts. Implemented retrieval-augmented generation with role-based access controls across 4 integrated systems. Built a model-routing layer with dynamic model selection and fallback policies. Demoed to 7 users at Sprint to Innovate 2026; pitched for the CIPS Manitoba Challenge.

Spatially Embedded Graphs

2024–Present

Python · Algorithms · Visualisation

Modelled Hamiltonian paths in ℝ² with research presented at GADA lab meetings. Generated 200+ randomised instances, built visualisation pipelines, and validated approximation behaviour on 200+ inputs.

Hurlbert–Kamat Property in Perfect m-ary Trees

2024

SageMath · Combinatorics

Investigated the HK-property in perfect m-ary trees using SageMath and combinatorial analysis. Implemented computational checks across 1,000+ cases to support conjecture validation.

.devHacks — "Most Chaotic Evil Hack"

2025

React.js · FastAPI · BERT · Supabase · TailwindCSS

Winner of 'Most Chaotic Evil Hack' among 20+ teams. Built a chatroom using websockets with a BERT + PyTorch text classification model that categorises messages (e.g. 'brainrot', 'nerdy', 'dad joke') and rewards users with in-game currency. Built the FastAPI backend with Supabase for persistent storage and message reactions.

CSSA CTF — 1st & 2nd Place

2025–2026

C · Python · Cryptography · Ghidra

1st place among 41 teams (2026) and 2nd place among 25+ teams (2025) at the annual CSSA Capture the Flag competition. Achieved 3rd place for Computational Sciences in the Summer Undergraduate Research Experience poster competition among 50+ student researchers.


Technical Skills

Languages JavaScript, TypeScript, Python, Java, C, C++, R, SageMath, BASH, SQL
Databases PostgreSQL, MongoDB, SQLite, MSSQL, Pinecone (Vector DB), Redis
Web Frameworks Next.js, React.js, FastAPI, Node.js, Astro
ML / Data PyTorch, PyTorch Geometric, NumPy, SciPy, Scikit-learn, MLFlow, Pandas
Cloud & DevOps AWS, Docker, Tmux
Developer Tools Git, Neovim, VS Code, IntelliJ, Figma, Jupyter
Specialised AI MCPs for Application Integration, Three.js, Beamer (LaTeX)