Jingran J.

Senior Data Scientist and Analytics Engineer, 7+ years in Canadian fintech and SaaS: KOHO, Wealthsimple, Samsara. Google Cloud Professional Data Engineer, AWS Certified Machine Learning, dbt Analytics Engineering. Applied Data Science, University of Michigan. I build reliable data and AI systems that help teams decide where to invest, which customers to retain and how to manage risk. My background spans Canadian fintech and SaaS, including KOHO, Wealthsimple and Samsara. I combine data science, analytics engineering and AI evaluation, from predictive models and attribution to dependable assessment pipelines. 📊 $1M saved a year: in-house real-time fraud detection on AWS SageMaker, replaced a paid vendor and lifted precision 17%. 📈 42% credit risk model performance lift: default probability model on LightGBM and XGBoost, 40,000 users approved for higher limits and $800K new revenue. 🎯 16.6% fewer misattributed conversions: unified marketing attribution across Meta, Google Ads, AppsFlyer, Amplitude and GA into SQL and dbt models. 🔄 40.4% re-engagement lift: RFM lifecycle segmentation that fed A/B retention campaigns. 💡 22% projected ROI gain: Bayesian media mix model in Google Meridian that found 10 to 15% of paid spend going to wrong channels. 🎮 ~10% Day-2 retention gap closed: K-Mode player segmentation across 2M+ users in iGaming, promo budget reallocated to the tiers that returned it. ⭐⭐⭐⭐⭐ 𝗪𝗛𝗔𝗧 𝗖𝗟𝗜𝗘𝗡𝗧𝗦 𝗦𝗔𝗬 "Lilian was exceptional as an NLP specialist on our engagement. She delivered a clear, implementation ready taxonomy review and refinement plan, brought deep rigor to error analysis and thresholds, and consistently elevated the quality of our decision layer thinking." - AI and NLP startup, model evaluation engagement "Lilian is a highly skilled statistician who was able to help implement a media mix model using Google Meridian. I highly recommend her for your data analytics tasks." - Marketing agency, MMM engagement 𝗪𝗛𝗔𝗧 𝗜 𝗗𝗢 🔹 Data Science and Machine Learning. Predictive models, propensity models, uplift models, recommendation systems. Python, scikit-learn, XGBoost, LightGBM, PyMC. 🔹 Analytics Engineering. dbt models, warehouse design in BigQuery, Snowflake and Redshift, ingestion pipelines, semantic layers a small team can maintain after I leave. 🔹 Marketing Analytics and Attribution. Multi-touch attribution, media mix modeling in Meridian and Robyn, incrementality testing, the SQL layer that makes numbers agree. 🔹 Product Analytics. Event tracking design, funnel analytics, cohort analytics and activation measurement in Amplitude, Mixpanel and GA4. 🔹 Retention and Lifecycle Analytics. Churn prediction models, RFM segmentation, LTV modeling and lifecycle campaign analytics. 🔹 Risk and Fraud Machine Learning. Credit scoring, default probability, real-time fraud detection on AWS SageMaker. 🔹 Business Intelligence and Dashboards. Executive reporting in Looker, Looker Studio, Power BI and Tableau. Data pipelines feeding boards that leaders actually read. 🔹 AI Implementation and Evaluation. Claude and OpenAI in production, RAG, structured tool use, evaluation frameworks, taxonomy design. 𝗥𝗘𝗖𝗘𝗡𝗧 𝗣𝗥𝗢𝗝𝗘𝗖𝗧𝗦 🔸 Unified marketing attribution and analytics layer for a Canadian fintech app across Meta, Google Ads, AppsFlyer, Amplitude and GA4. 🔸 Real-time credit risk and fraud detection systems for a Series D fintech lender on AWS SageMaker. 🔸 Player value-tier segmentation across 2M+ users for an international iGaming operator, closing a 10-point day-2 retention gap. 🔸 Bayesian media mix model in Google Meridian for a high-spend marketing agency's client book. 🔸 Live Claude LLM pipeline for interview scoring and remote hiring fraud detection at a workforce intelligence SaaS. 🔸 NLP harm classification decision layer for an AI startup, 70% fewer false positives with BERT fallback. 𝗛𝗢𝗪 𝗜 𝗪𝗢𝗥𝗞 1. Read-only assessment first. I look at what exists, what is authoritative and where the analytics break, before proposing anything. 2. Agree the decision the work supports. Every model and every analytics build gets pointed at a decision someone is actually making. 3. Build in the smallest maintainable form. You own the code, the models and the analytics. 4. Hand over with documentation and a walkthrough, so the analytics and the machine learning survive after I leave. 𝗧𝗢𝗢𝗟𝗦 Python, SQL, R, dbt, BigQuery, Snowflake, Redshift, Databricks, Amazon Web Services, SageMaker, Airflow, Fivetran, Segment, Hightouch. Looker, Looker Studio, Power BI, Tableau, Metabase. Amplitude, Mixpanel, Google Analytics 4, Heap. Claude, OpenAI, Anthropic, RAG pipelines, vector search, prompt engineering. 📩 Send me a note about your current data setup and the business decision you need to support. I’ll help clarify the scope, priorities and best starting point.