Internship Experience

Shopify
Shopify

Machine Learning Engineer Intern

Sep 2026 - Present
Montreal, QC, Canada
  • Benchmarked GPU inference for product embedding models with TensorRT across 50 batching configurations, pinpointing bottlenecks at 600 requests/sec and selecting the production serving setup.
  • Deployed a distributed NVIDIA Triton inference service that compresses product embeddings into compact semantic IDs, powering generative search and recommendations for 3 downstream teams.
  • Automated ML validation pipelines in Python on GCP, verifying 60,000+ embeddings across deployed models at 0.999+ cosine similarity during a zero-downtime runtime upgrade.
  • Provisioned BigQuery infrastructure as code with Terraform across staging and production, enabling large-scale embedding backfills into new production tables.
  • Converted PyTorch embedding models to ONNX and compiled TensorRT engines for GPU serving on NVIDIA Triton.
  • Containerized the inference service with Docker and rolled it out on Google Kubernetes Engine, with Prometheus and Grafana dashboards tracking latency, throughput, and GPU memory.
  • Served semantic IDs to downstream search and recommendation services over Triton's gRPC endpoints, tuning timeouts and concurrency limits for steady throughput.
IBM
IBM

Forward Deployed Engineer Intern

May 2026 - Aug 2026
Toronto, ON, Canada
  • Embedded with enterprise clients to automate 15+ document workflows with multi-agent systems on watsonx.ai that extract, validate, and route invoices via RAG and A2A, reducing manual processing time by 60%.
  • Drove 4 client proofs of concept from discovery to working demo, building FastAPI connectors that let AI agents read and write to SAP, ServiceNow, and SharePoint, with 3 advancing to production evaluation.
  • Hardened client agents with PII redaction and prompt-injection filtering, packaged as a reusable guardrail layer adopted by 3 other IBM teams.
  • Delivered 6 technical workshops and hands-on labs to 200+ engineers on building agentic RAG pipelines with watsonx.ai.
  • Authored evaluation harnesses that test agent accuracy and tool-call reliability against client test sets before handoff.
  • Introduced hybrid semantic and keyword search over Elasticsearch for document-processing agents.
  • Parallelized independent agent tool calls with async Python and tuned retrieval for client assistants, bringing end-to-end response time down from 12s to 4s.
Autodesk
Autodesk

Machine Learning Engineer Intern

Jan 2026 - Apr 2026
Montreal, QC, Canada
  • Designed a Python agent framework on Amazon Bedrock, enabling 6 teams to ship 50+ in-product AI agents.
  • Standardized agent tool access behind shared MCP servers, shrinking new tool integration time by 80% and replacing duplicated per-team integration code with one maintained service.
  • Implemented token streaming over Server-Sent Events for agent responses, cutting time-to-first-token by 65% across live agent workflows.
  • Launched a token quota service and dashboard for 50+ AI agents, lowering LLM spend by 25%.
  • Structured the framework around the Amazon Strands SDK with a modular architecture, letting teams add tools and agents without rewriting core logic.
  • Stood up a RAG pipeline on AWS using Amazon S3 for document and vector storage, enabling 6 product teams to ingest 50,000+ chunks and improve relevant document retrieval by 55%.
  • Released services to DEV, STG, and PRD with Spinnaker, Jenkins, and Terraform, using infrastructure as code for consistent promotion between environments.
  • Wrote unit and integration tests with PyTest for the agent framework, mocking Bedrock clients to catch regressions before release.
UKG
UKG

AI Software Engineer Intern

May 2025 - Aug 2025
Montreal, QC, Canada
  • Shipped 15+ LangGraph agents in Python resolving PTO, scheduling, and payroll requests for 10K+ users.
  • Engineered hybrid-search RAG over HR policy documents in MongoDB and SQL, raising recall by 40%.
  • Integrated a Redis semantic cache that matches new queries to past ones by embedding similarity, trimming token usage by 60% and serving cached answers 23x faster than full LLM calls.
  • Exposed agent tools as FastAPI services, containerized with Docker and pushed through GitHub Actions CI/CD pipelines.
  • Instrumented agents with LangSmith tracing and structured logging to debug failures across multi-step workflows.
  • Assembled an offline evaluation set of labeled queries to regression-test prompt and retrieval changes before release.
Biomomentum Inc.
Biomomentum Inc.

Software Developer Intern

May 2024 - Aug 2024
Montreal, QC, Canada
  • Engineered a real-time signal processing pipeline in Python for biomechanical testing hardware, reducing latency by 80% via multithreading and algorithmic optimization.
  • Automated QA workflows in Python, cutting manual testing effort by 25% and enabling CI/CD integration.
  • Optimized custom VHDL code for Spartan FPGAs, increasing acquisition speed by 15%.
  • Programmed FPGA interfaces for ADS1298 ADCs and LSM9DS1 IMU with custom SPI, I2C, UART modules.

Extracurricular & Leadership

McGill Robotics

System Manager

Sep 2024 - April 2026
Montreal, QC, Canada
  • Led a 15-member software team, overseeing system architecture and improving task completion by 30%.
  • Assigned and tracked 15+ monthly tasks across the team.

Software Developer

Sep 2023 - April 2026
Montreal, QC, Canada
  • Built real-time coordination scripts and debugging tools in Python, C, and C++ enabling efficient control of 20+ hardware components.
  • Developed STM32 firmware in C for CAN communication reducing signal delay from 10ms to 3ms.
  • Engineered custom ESP32 remote kill switch in C++ reducing response time from 5s to <1s.
  • Designed Python scripts on NVIDIA Jetson GPU coordinating 20+ hardware components.
McGill AI Lab
McGill AI Lab

Machine Learning Researcher

Sep 2024 - Dec 2025
Montreal, QC, Canada
  • Designed and trained deep learning models in PyTorch and NumPy for MNIST digit classification, achieving 98%+ accuracy via dropout, batch normalization, and hyperparameter optimization.
  • Identified key performance bottlenecks using Matplotlib visualizations, improving model convergence rates.
  • Implemented data preprocessing pipelines with NumPy and scikit-learn to handle 60,000+ training samples, ensuring robust feature scaling and cleaner inputs for model training.
  • Experimented with multiple neural network architectures (CNNs, fully connected layers, regularization strategies) to benchmark trade-offs between accuracy and efficiency.
Marianopolis College
Marianopolis College

Teaching Assistant

Jan 2024 - Dec 2024
Montreal, QC, Canada
  • Graded assessments and assisted in delivering Professor Baharak Fatholahzadeh's college Physics courses for 150+ students.
  • [Fall 2024]
  • Waves, Light, and Modern Physics: Oscillatory Motion, Wave motion, Sound Waves, Superposition and standing waves, Ray Optics, Image Formation, Interference of waves, Diffraction, Relativity, Quantum Physics, Nuclear Decay, The Optics of the Compound Microscope.
  • [Winter 2024]
  • Electricity and Magnetism: Electric Fields, Electric Potential, Capacitance and dielectrics, Direct current circuits, Magnetic Fields and Sources, Faraday's Law and Inductance, Continuous Distribution of Charge.

Education

McGill University

McGill University

B.Eng. Electrical Engineering, Minor in Applied AI

2023 - 2027
Montreal, QC, Canada

GPA: 3.8 / 4.0

Marianopolis College

Marianopolis College

Health Sciences

2021 - 2023
Montreal, QC, Canada

Dean's List both years enrolled

Skills

Languages

PythonCC++JavaSQL

Libraries / Frameworks

LangChainLangGraphAmazon StrandsAmazon BedrockPyTorchTensorFlowNumPyPandasMpMatplotlibFastAPIFlaskSQLAlchemyPyMongo

Developer Tools

AWSGCPDockerKubernetesTerraformSpinnakerSplunkGrafanaGitGitHubMySQLMongoDB