Skip to content
View yanischalel's full-sized avatar

Block or report yanischalel

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
yanischalel/README.md

Hi, I'm Yanis

AI engineer working on LLM systems in production — agent design, tool and data integration, evaluation, and deployment.

AI Engineer on Bell's Applied AI team, building and deploying agents on Google Cloud with Gemini Enterprise and Antigravity. ML research assistant at GEODES, the software engineering lab at Université de Montréal. Member of the Cohere Labs Open Science Community.

Recent work

Crisis-detection guardrail — 1st of 77 teams at the Mila × Bell × BuzzHPC × Kids Help Phone AI Safety Hackathon. Real-time conversation safety classification for youth mental health support: a fine-tuned multilingual BERT (mmBERT) feeding a Cohere LLM judge through a five-question chain-of-thought framework, with the classifier score injected into the prompt as a prior rather than ensembled. F1 0.906, recall 0.969 — the highest recall in the top 10.

Automated code review — research at GEODES, advised by Prof. Houari Sahraoui. Fine-tuning small open-source code models with QLoRA / PEFT to close the quality gap with large proprietary models, plus an evaluation methodology that benchmarks generated reviews against human reviewers.

Dynamic pricing in Paris — ~60,000 Airbnb listings assembled from historical snapshots and a custom multithreaded scraper, testing whether explicit temporal features improve price prediction. Random Forest, R² 0.525 with uncapped luxury listings retained.

Talks

  • AI Tinkerers Montreal — recall-first safety classifiers for high-stakes settings
  • CRIM 40th Anniversary Conference — AI safety for LLMs

Stack

Agents & LLMs — Python, Gemini, Gemini Enterprise, Antigravity, MCP, LLM-as-judge evaluation, chain-of-thought prompting

Model training — PyTorch, Hugging Face Transformers, QLoRA / PEFT, BERT / mmBERT

Cloud — Google Cloud Platform (Vertex AI, BigQuery), AWS

Backend — Go, Java, Node.js

Frontend — TypeScript, React

Infra — Kafka, Docker, Kubernetes, CI/CD, Prometheus / Grafana

Previously

GoTo (2024–2026) — three internships into an engineering role. Go/Kafka event pipelines, React/TypeScript frontend architecture, Java data pipelines.

Ubisoft GameLab 2026 — gameplay programmer, Unreal Engine 5 and C++.

BSc Computer Science, Université de Montréal (2022–2026), with two exchange semesters at the University of Hong Kong (W2025) and Sorbonne Université (F2025).

Contact

LinkedIn · yanis.chalel@umontreal.ca

Always happy to talk about LLM evaluation, agent design, AI safety, or open-science collaborations.

Pinned Loading

  1. multilingual-guardrail-mila-hackathon multilingual-guardrail-mila-hackathon Public

    Stacked multilingual guardrail (mmBERT + Cohere c4ai LLM judge) for crisis detection in youth chat. 1st of 77 teams at the Bell x Mila x BuzzHPC x Kids Help Phone AI Safety Hackathon. F1 0.906, rec…

    Python 1

  2. Airbnb-Price-Prediction Airbnb-Price-Prediction Public

    Dynamic pricing of Airbnb listings in Paris using a hybrid dataset. Compares static models vs. temporally enriched models (month, seasonality) with Random Forests, combining InsideAirbnb data and c…

    Jupyter Notebook

  3. Kerwan0916/ConcoursUbisoft2026 Kerwan0916/ConcoursUbisoft2026 Public

    C++