Om Balar

AI Software Engineering Trainer

DataAnnotation

2026-06 - Present

Tested AI coding assistants across 5 projects by writing specific prompts to find errors, expose broken logic, and identify fake answers (hallucinations). Improved AI training data by reviewing AI-generated code in large open-source projects and writing clear reports on system design mistakes. Built safe, isolated testing environments using Docker to run and check code written by AI models.

Software Engineer

Omello

2025-12 - Present

Shrank the app size from 900MB to 250MB by cleaning up the code and removing unused packages. Built a system using GitHub Gists to update AI model names automatically, preventing the app from breaking when Google Vertex AI or Groq update their APIs. Built an AI pipeline faster using Claude Code and Cursor that transcribes live audio and uses context-aware search (RAG) to create study notes in real time. Fine-tuned a custom PyTorch neural network to run efficiently on local hardware while strictly staying under a 500MB memory limit.

Software Developer

Environment and Climate Change Canada

2025-05 - Present

Cut data processing time from 90 seconds to 10 seconds by adding database caching to reduce repeated searches. Built a reliable Python and SQL data pipeline to process ~1.5 million weather forecast files smoothly without crashing. Proposed and built a custom formatting toolbar for the internal text editor, making it easier for meteorologists to use. Sped up software releases and cut automated testing time from 12 hours to 7 hours by improving code quality and setting up pipelines with Docker and Cypress.

Software Engineer

Distributed Job Search Platform

2025-01 - 2025-04

Built a reliable job search website using React and Node.js that lets users search through large amounts of data quickly. Kept the website online during heavy traffic by running 3 worker nodes using Docker and Kubernetes on Google Cloud Platform (GCP).

Software Engineer

Reviewer.ly

2024-05 - 2024-08

Cut live software bugs by 30% by building backend API services with Java and gRPC, and writing thorough automated tests with Jest and Mockito. Sped up the application by moving heavy image processing tasks from the user's browser to scalable backend servers. Helped the team fix problems faster by building an automated email notification system to alert engineers of platform issues.