Thomas SemczyszynSoftware Engineer · AI & Agents

Toronto, ON·Jan. 2026 – Oct. 2026

AI driven financial intelligence startup

• Founding Engineer

  WHAT I DID

  The data side was mine: the ingestion and the storage that everything else reads off. I shipped features end to end, ran the deploys myself, and picked up whatever broke after. This experience made me confident that I can build anything.

  • Built the serverless ingestion pipeline on Lambda, SQS and EventBridge. It pulls hourly options data for over four thousand tickers and runs a hundred odd calculations over it into MongoDB.
  • Backfilled five years of daily stock and options history with parallel, batched requests, held under the API rate limit of three thousand a minute.
  • Built the provider layer the agents run on, with OpenAI, Gemini and Perplexity behind one interface. Each keeps a thread going through its own response id rather than by resending the conversation every turn.
  • Kept a registry of about two dozen models carrying per million token pricing and what each one supports, so a call could be costed and routed before it was made.
  • Wrote the layer that reads a written investment thesis and turns it into structured intent and follow up questions in a single pass.
  • Set up deploys to ECS Fargate on merge to main through GitHub Actions.

Toronto, ON·May 2025 – Jul. 2026

Mental health software company

• Software EngineerSep. 2025 – Jul. 2026

  WHAT I DID

  Stayed on full time after the internship, on the backend of a mental health platform. Mostly features end to end: the endpoints, the tables underneath them, and the performance work when something needed to be faster.

  • Messaging was slow enough to be noticeable, so I went and rewrote the query behind it. It came back about ten times faster.
  • Shipped REST endpoints with pagination and filtering, and the database tables for a new product feature, custom consent forms.
  • Moved production logging onto contextLogger so it all lands in Elastic with context attached.
  • Wrote up how to wire the model built by the external ML team into the backend, for the rest of the team to build on.
  • Stayed on the product team and the awards committee after converting from the internship.
• Software Engineering InternMay 2025 – Aug. 2025

  WHAT I DID

  Started here as an intern, and put real effort into getting involved past the work I was handed. I got a feel for how a production system actually operates, and how the whole development cycle works, from the back and forth with the QA and frontend team.

  • Wrote and tested migration scripts, with check constraints to keep the data honest.
  • Ran end to end tests in Docker across platforms, with AVA covering unit and integration.
  • Extended a local Keycloak registration token with custom user attributes and wired them through the API.
  • Investigated and patched a set of high risk security vulnerabilities.
  • Joined one of the product teams, helping define requirements and review outputs while they worked with an external ML team.
  • Sat on the awards committee.

New York City, remote·Sep. 2024 – Dec. 2024

Financial technology startup

• Software Engineering Intern

  WHAT I DID

  My first proper engineering job, on a small fintech team in New York. I lived in the Django backend, implementing new REST API endpoints, conducting automated testing, and integrating services like Sentry and Datadog for monitoring.

  • Built REST endpoints and fixed production bugs as they came in.
  • Wrote Selenium tests for the user flows nobody wanted to click through by hand every release.
  • Set up Sentry and Datadog so we could see errors and latency instead of guessing at them.

Remote, freelance·May 2024 – Feb. 2025

Freelance LLM training and data refinement

• AI Trainer

  WHAT I DID

  Freelance work training large language models, picked up alongside school. The whole job was to be the person who notices when a model is confidently wrong, which is a habit I have not been able to switch off since.

  • Wrote coding prompts across about a dozen languages for instruction tuning, on models including BART and Gemini.
  • Compared competing responses side by side and ranked them on correctness, instruction following, and whether the model had actually understood what was asked.
  • Reviewed output for hallucinations, and corrected several hundred answers, mostly logic that looked right and was not.

Nov. 2023

I never thought I was a runner. Six months before this I had never run more than 3 kilometres. An official marathon is 42.2 km, but I kept going and finished at 46. Definitely my proudest moment.

• Completed46 km

Toronto, ON·Nov. 2022

I randomly took an introduction to computer science course as a biology major because I needed an elective, and I'd always hear people say that coding's an important skill and everyone should know how to code a bit. I found the course to be quite challenging, and I even remember struggling with simple things like "for loops", but I loved the challenging aspect of it, and found it to be quite mentally stimulating. After this single course I decided that I was going to start studying computer science :)

• It started withprint("Hello world")

Toronto, ON·Sep. 2020 – Apr. 2025

Bachelor of Science

• Computer Science & Statistics

Relevant Coursework:Software Development, Databases, Machine Learning, Data Structures and Algorithms, Object Oriented Programming, Systems & Tool Programming, Computer Organization, Data Analysis