Hi, I'm

Shubhnoor Gill

I design and build AI systems, data pipelines, and automation that companies actually use. Based in Toronto - open to full-time roles, consulting, research partnerships, and conversations worth having.

Full-time RolesConsultingResearchPartnerships

Impact

Numbers that matter

I bridge technical AI and real business outcomes - from building LLM pipelines and RAG systems to running executive workshops and driving adoption across organizations. Currently leading data and AI at Paul Street in Toronto, with experience spanning SaaS, startups, marketing, consulting, and academia.

0%

Reduction in manual ops time

AI automation at Paul Street

0%

Reduction in reporting time

ETL and dashboard automation

0%

Faster onboarding

LLM contract analysis

Career

The journey

Work

Projects

MLAI/LLM

AI-Driven Credit Card Fraud Detection

Built a fraud detection system using advanced data balancing techniques and ensemble/deep learning models to handle extreme class imbalance (0.17% fraud rate). Achieved F1-scores above 0.9 using CatBoost, XGBoost, and neural networks - minimizing false positives and operational risk.

PythonCatBoostXGBoostScikit-learn
AI/LLMResearch

LLM Fine-Tuning and Evaluation System

Built a data science and AI assistant using web scraping, data preprocessing, and model fine-tuning. Fine-tuned Google Gemma 2B and 7B LLMs using PEFT techniques (LoRA, QLoRA). Implemented LLM evaluation frameworks using LangChain string evaluators and Prometheus templates.

GemmaLoRAQLoRALangChainPython
MLMarketingAnalytics

Executive AI Marketing Strategy Optimizer

AI-driven marketing analytics solution for campaign performance and executive decision-making. Applied predictive modeling and k-means customer segmentation, resulting in 1.8x ROI lift and 60% reduction in campaign spend. Descriptive analytics delivered via Power BI dashboards.

PythonXGBoostK-MeansPower BI
AI/LLMAutomationProfessional

LLM Contract Analysis System

End-to-end GenAI solution using AWS and RAG to automate structured data extraction from unstructured legal contracts. Reduced client onboarding time by 40% and operational costs by 25% at PandoPartner.

LLMsRAGAWSPython
AnalyticsMarketingProfessional

MTA and MMM Attribution Dashboards

Executive-facing multi-touch attribution and marketing mix modeling dashboards for partner network analysis. Attribution modeling across 5+ partner networks contributed to a 15% lift in net-new customer acquisition.

BigQueryLookerSQLPython
Computer VisionResearch

Hand Gesture Interactive Gaming

Real-time computer vision system that uses hand gestures to control virtual games. Zero controller needed - built using MediaPipe for hand landmark detection.

PythonOpenCVMediaPipe

Expertise

Skills

AI and Machine Learning

LLMs (GPT, Claude, Gemini)RAG and Vector DatabasesNLPAI Agents and MCPFine-Tuning (LoRA, QLoRA)Predictive AnalyticsData ScienceResponsible AI and Data Privacy

Programming and Data Engineering

Python (Pandas, NumPy)SQLETL PipelinesREST APIsFeature EngineeringData CleaningDagsterCloud: BigQuery and GCP

Analytics and BI

Looker StudioPower BIGoogle Analytics 4MTA and MMMFunnel and Customer Journey AnalysisSEO and Campaign Management

Tools and Platforms

GitHubPostmann8nMakeNotionAirtableAsanaLinearReplitConfluence

Contact

Let's talk

Open to full-time roles, consulting engagements, research collaborations, and partnerships. Toronto-based, available globally.