Hashim Jama

I build software and work with data. Most of my projects pair a dependable backend with a model or an analysis on top. I enjoy turning raw data into something people use.

Open to new roles
Previously a software engineering intern at myAIpathway

See my work Get in touch

About

I studied Computer Science at Wilfrid Laurier University. During my internship, I connected large language models to a Node.js backend I built. Outside of work, I’ve built a live study app with Next.js and Supabase. I’ve also implemented a Transformer from scratch and built a machine learning model to forecast Canadian housing prices.

Whatever I’m building, I want the data well organized, the code tested and the system easy to run. Most of all, I want the results to make sense to the people who rely on them.

Languages
Python, Java, TypeScript, JavaScript, SQL, C#, C++, C
Backend & APIs
Node.js, Express, FastAPI, Django, Spring Boot, REST APIs, WebSockets
Frontend
React, Next.js, Astro, Tailwind CSS, HTML/CSS, AngularJS
Data & analytics
PostgreSQL, MySQL, MongoDB, Redis, pandas, NumPy, Matplotlib, Seaborn, Tableau, Power BI, Streamlit
AI & ML
LLM integration, RAG, NLP, TensorFlow, scikit-learn
Deployment & tooling
Docker, CI/CD pipelines, Git, Vercel, Supabase, GitHub Pages, Locust

Experience & education

Software Engineering Intern · myAIpathway Inc.

May to December 2025 · Remote

  • Designed an indexed MySQL schema that made log retrieval 40% faster, then added a product cache that reduced overhead by 60%.
  • Built the Node.js and Express backend for an AI support assistant that kept track of long conversations and wrote its own prompts as each one unfolded.
  • Added retry logic so LLM calls recovered from rate limits and created a stateless API to track when conversations were handed off.
  • Wrote a context engine that counted tokens and trimmed older messages so every prompt fit within the model's limit.

B.Sc. Computer Science · Wilfrid Laurier University

2022 to 2026

Selected work

AI powered feed engine

This FastAPI service builds a personalized feed for every user. Each post is scored on how well it matches the user's interests, its quality, how recent it is and how often it gets clicked. The feed then mixes in other topics so no single subject takes over. Feeds are cached in Redis and refreshed whenever a user clicks something. New posts reach interested users right away over WebSockets. The whole stack runs in Docker Compose alongside Postgres and Redis. On every push, GitHub Actions lints the code, runs the tests against real databases and starts the full stack as a smoke test.

  • FastAPI
  • PostgreSQL
  • Redis
  • WebSockets
  • Docker
  • GitHub Actions

Source on GitHub for AI powered feed engine, opens in a new tab CI runs for AI powered feed engine, opens in a new tab

StudySync

StudySync is a Next.js and Supabase app that helps students find study partners and groups. A Postgres function scores every possible match on shared courses, overlapping study times and learning style, so the best fits appear first. Group chats and shared files update live through Supabase Realtime. Database triggers send a notification when someone joins your group or posts a message. Row level security makes sure students only see the groups and conversations they belong to.

  • Next.js
  • TypeScript
  • Supabase
  • Postgres
  • Realtime

Live demo of StudySync, opens in a new tab Source on GitHub for StudySync, opens in a new tab

Neural machine translator

I built this English to French translator from scratch in TensorFlow, following the design of the original Transformer paper. It uses eight attention heads, positional encoding, masking and the paper's warmup learning rate schedule. The model has six encoder layers and six decoder layers, which I trained on the WMT14 dataset using subword tokenizers.

  • TensorFlow
  • Python
  • NLP
Source on GitHub for Neural machine translator, opens in a new tab

Canadian property predictor

This machine learning pipeline forecasts Canadian house prices by province using Statistics Canada data from 1990 onward. I compared a Random Forest with a Gradient Boosting model by training on 1990 to 2017 and testing on the years since. The Random Forest won with an R² of 0.95 and predicts each year's growth, which lets a Streamlit app project prices from 2026 as far as 2100 under different economic scenarios. Matplotlib and Seaborn charts show how both models performed.

  • scikit-learn
  • pandas
  • NumPy
  • Matplotlib
  • Seaborn
  • Streamlit
Source on GitHub for Canadian property predictor, opens in a new tab

Minimax chess engine

I built this chess game in Pygame to follow the full rules, including castling and en passant. It also tracks pinned pieces, so it never allows a move that would leave a king in check. The computer opponent looks three moves ahead using negamax with alpha beta pruning. Its search runs in a separate process so the board keeps animating while it thinks.

  • Python
  • Pygame
  • multiprocessing
Source on GitHub for Minimax chess engine, opens in a new tab

AdBlocker Pro

This Chrome extension filters network requests to stop ads, trackers and social widgets before they load. If a page slips an ad in later, the extension spots it and removes it right away. The popup shows how many items it has blocked today and in total. From there you can also pause blocking or allow sites you trust.

  • JavaScript
  • Chrome MV3
  • declarativeNetRequest
Source on GitHub for AdBlocker Pro, opens in a new tab

Contact

I'm looking for my next role in software or data. Email is the quickest way to reach me.