I take messy, real-world data and turn it into pipelines, models and dashboards that answer one question well: what should we do next? I care about clean data, honest numbers, and reports people actually open.
Data Analyst Intern, NCPLIndia, 2024. Dashboards and report automation
Northeastern UniversityMaster of Professional Studies, Data Analytics, Toronto
Capstone with Nytia LabsHealthcare risk prediction on Google Cloud, Spring 2026
NowBuilding cloud data projects and open to analyst roles across Canada
Projects
End-to-end work, from raw data to a decision. Each colour line marks the main tool behind it.
Featured project
TTC Delay Analytics: a fully automated pipeline on live Toronto transit data
Where do Toronto’s subway, bus and streetcar delays really come from, and which stations get hit the hardest? I built a pipeline that answers that every single day without me touching it.
Pulls fresh delay data from the City of Toronto Open Data API with a Python Azure Function on a daily timer, saved to date-partitioned Blob Storage.
A scheduled Snowflake Task and stored procedure rebuild the warehouse daily: raw JSON, cleaned staging views, then a star schema (fact_delays, dim_delay_code, dim_date).
The dashboard shows buses carry about 80% of 2.57M delay minutes across 164K incidents, bus delays average 21 minutes against 2.7 for subway, diversions are the single biggest cause, and Kennedy Station is the worst hotspot.
Power BI dashboard with KPI cards, top delay causes, monthly trends, mode comparison and worst stations. Built in Import mode so the report keeps working even if the cloud trial ends.
Nytia Health: Complication Risk Prediction on Google Cloud
Capstone for Nytia Labs, an Ontario digital health company (Spring 2026). Which patients with chronic conditions are heading toward complications, so care teams can step in early?
Our nine-person team built an end-to-end pipeline on Google Cloud: BigQuery bronze, silver and gold layers, Cloud Run jobs for ingestion, features and training, and a live FastAPI prediction endpoint behind a dashboard.
Engineered compound features, like how many wellness areas are declining for each user. These correlated with risk at 0.69 while single areas were weak, showing risk comes from decline across many areas at once.
Validated the approach on the sponsor’s full 430 million records in BigQuery ML, where the XGBoost model reached 0.945 accuracy and 0.992 AUC.
Spotted that the full dataset was synthetically generated and reported it to the sponsor, so the results are read correctly: the pipeline is proven and ready for real clinical outcome labels.
Which borrowers are likely to default, and what drives that risk? A scoring model plus a dashboard a credit team could use, built on 150,000+ customer records.
Cleaned the data, engineered 6 new risk features, and compared three models. XGBoost won with a 0.86 ROC-AUC, ahead of logistic regression and random forest.
Found late payment history is the strongest signal: borrowers with any late payment defaulted at 22.3%, against 2.7% for those without. The model’s risk tiers separate cleanly too: 93.7% of “very high risk” customers defaulted, against 2.9% of “very low risk”.
Loaded scores into a 5-table PostgreSQL schema, automated daily scoring with a Python pipeline, and published a Tableau dashboard of risk by segment.
How has traffic across Toronto changed over three decades, and where is it heaviest? An automated pipeline over 32 years of City of Toronto counts (1993 to 2025).
Built a Python ETL that pulls from the City of Toronto Open Data API into a SQLite warehouse, then wrote 30+ SQL queries on 14,000+ traffic readings.
Showed traffic fell about 30% during 2020 to 2021 and has been recovering since, and identified Highway 401 as the busiest corridor at roughly 110,800 vehicles a day. Weekend counts averaged 8,817 a day against 6,656 on weekdays.
Published a Tableau dashboard with a 32-year trend, top 10 busiest locations and a geographic heat map.
Where is a hospital under strain, and does staffing actually affect patient satisfaction?
Three dashboard pages. Select one to enlarge.
Found and fixed heavy duplication in the source data, built a calendar table, and wrote twelve analytical SQL queries.
Showed 40% of patients came back for repeat visits and flagged an unusually high readmission rate in Physical Medicine and Rehabilitation.
Found no statistical link between staff ratios and satisfaction, which shifts the focus to other causes. Delivered a three-page Power BI dashboard, report and slide deck.
MySQLPower BIDAXStar schema
WarmeHands Business Performance Dashboard
An end-to-end Power BI build where the hardest part was deciding which data to trust.
Checked data quality in Python (openpyxl) before import, then cleaned it in Power Query and modelled a star schema.
Dropped an unreliable country reference sheet and reclassified cost types after an early formula produced an impossible margin.
Built DAX measures and a profitability dashboard: 503.9K in revenue at a 42.2% margin across 305 orders, with the UK as the top market.
Power BIPower QueryDAXPython
No projects with that tool yet. Pick another filter.
Experience & education
Data Analyst, Capstone Project, Nytia Labs
Toronto, Spring 2026, nine-person team
Helped build a system for Nytia Labs that sorts patients with chronic conditions into low, moderate and high complication risk, deployed on Google Cloud.
Owned feature engineering: built the compound decline features that turned out to be the strongest risk signal (0.69 correlation).
Validated the model on the sponsor’s full 430-million-record dataset in BigQuery ML, reaching 0.992 AUC.
Data Analyst Intern, NCPL
India, January to December 2024
Built dashboards tracking how clients’ website features were used and which visitors became customers, with per-user profile views.
Automated a daily report, cutting it from about an hour to 15 minutes.
Master of Professional Studies, Data Analytics
Northeastern University, Toronto campus
Toolkit
Every tool here links to the project where I used it.
Wrote a stored procedure and a scheduled Snowflake Task that rebuild the TTC warehouse every day, plus Streams and Tasks for incremental loads in the retail warehouse.
I’m open to Data Analyst, AI Data Analyst and Credit Risk Analyst roles anywhere in Canada, and to freelance projects. I have a driver’s licence and a car, and I’m happy to relocate.