Nawaf Mahmood

Toronto. Computer Science co-op at TMU, class of 2028.

Toronto. Computer Science co-op at TMU, class of 2028.

I'm Nawaf. I build AI tools at Sanofi and ship my own products on the side.

Right now that means Newton, the platform about 700 people at Sanofi use to talk to their data, and TraceReview, the tool I built so we can see what its agents are actually doing. Outside work I run VisualBuild, live on the App Store, and TMU Pulse.

Nawaf Mahmood

Day job

Full-stack developer at Sanofi, on the CX Data & AI and Commercial M&S teams.

On the side

VisualBuild and TMU Pulse. Both live, both with real users.

Also

Research assistant to Prof. E. Harley at TMU, and working toward AWS Solutions Architect.

Full-Stack Software Developer. May 2026 to now. Toronto.

Sanofi

I work across two teams, CX Data & AI and Commercial M&S. Day to day that is React and Python on top of Snowflake, running on AWS, with Power BI for reporting. I also help ship web releases, which move from dev through UAT to production.

~700
people using Newton
~500
messages a day
5
teams on the platform
15+
data sources unified for CX

Newton, Talk to Data

Sanofi's company-wide platform for asking data a question. Someone asks. Snowflake agents work out the SQL, run it, and come back with results and charts. I ship full-stack features across the platform.

  1. Chat UI and prompt libraryWhat people see
  2. Access via Microsoft GraphWho can see which data
  3. Agents that write the SQLQuestion in, query out
  4. SnowflakeWhere the data lives

Shipped so far: Microsoft Graph access groups, Prompt library, Agent selection, File upload, Skills, Data refresh.

TraceReview, led end to end

One screen that shows what an agent actually did. I scoped it with the agent owners, built it in React and Python, and shipped v1 in two weeks. For every run it lays out the agent's reasoning, the SQL it wrote, the results, and the charts.

  1. 1Read the question and pick the right table.
  2. 2Check which columns answer it.
  3. 3Write the query, run it, then chart the rows.

Run timeline

Illustrative mock. Not a real screenshot.
  • ~60% less time spent debugging agents
  • ~5 production incidents caught
  • Started as a Newton-only tool. Now any team can use it: 2 teams do today, and RUSH is joining
  • Helped migrate it onto AWS

Web releases

Releases move from dev, through UAT, to prod on a cycle, and a different group of people runs each one. I help with that.

CX Data & AI

15+ sources, one database. Automated ETL pipelines pull SharePoint, Google Analytics, and internal systems into a single database behind the team's React and Power BI dashboards.

  • React
  • Python
  • Snowflake
  • AWS
  • Power BI
  • SQL
  • Microsoft Graph
  • SharePoint
  • Google Analytics

Live on the iOS App Store. April 2026 to now.

VisualBuild

A contractor types in an address. About thirty seconds later they are showing the homeowner their own house with the renovation already on it.

visualbuild.caApp StoreDetails

AfterBefore
Drag, or use the arrow keys
  1. 01Address in. No photo upload. The contractor enters a street address.
  2. 02Three angles. The pipeline pulls three views from the Google Street View API.
  3. 03Auto-clean. Wires, cars, and neighbouring houses come out of the shot.
  4. 04Render. The renovation is rendered in about 30 seconds, across 21 material categories.
  • Row-Level Security and shared JWT auth across web and iOS
  • AI keys stay server-side
  • A backup model has held 500+ renders near 100% success through provider outages
  • 2 businesses onboarded in early access
  • Next.js
  • TypeScript
  • Tailwind CSS
  • Supabase
  • React Native (Expo)
  • Gemini API
  • OpenAI API
  • Stripe

Live for TMU students. June 2025 to now.

TMU Pulse

An academic planning platform for TMU students. GPA tracking, schedule building, course discovery, and degree planning in one place, with Rate My Professors built in.

tmupulse.caDetails

500
sign-ups
30,000+
visits
~100
visits a day
3,700+
course records indexed
  • TMU Pulse course catalogue

    Course catalogue

  • TMU Pulse home page

    Home

  • TMU Pulse schedule builder

    Schedule builder

A course catalogue that stays fast.

A Python and PostgreSQL pipeline scrapes, normalizes, and indexes 3,700+ course records across 100+ departments. Pages load in under a second.

  1. Scrape
  2. Normalize
  3. Index
  4. Sub-second loads
  • GPA tracker
  • Schedule builder
  • Course discovery
  • Degree planner
  • Python
  • Next.js
  • TypeScript
  • PostgreSQL
  • Supabase
  • React
  • Tailwind CSS

Undergraduate Research Assistant, TMU. June 2026 to now.

Research with Prof. E. Harley

The lab edits genes and needs to know whether each edit actually worked. I built the pipeline that checks, straight from raw DNA-sequencing data.

Details

  1. Before

    Checked by hand

    Results were verified by manual inspection, batch by batch.

  2. The pipeline

    Four tools, one run

    Python chaining fastp, SPAdes, BLAST, and minimap2 across 6 experimental batches. Assemblies validated at 99.8% accuracy.

  3. What it found

    The edit had silently failed

    I traced an unintended copy elsewhere in the genome, and the lab redirected its next round of experiments.

ReadsIntended site, edit not thereUnintended copy,found elsewhere in the genome
Schematic, not real data.
  • 01fastp Read cleanup
  • 02SPAdes Assembly
  • 03BLAST Sequence search
  • 04minimap2 Alignment

1st place. Sanofi x Snowflake Hackathon 2026.

RepPal

Under Canada's CASL law, a rep gets one shot at asking a physician for consent. RepPal recommends which physicians to approach, and how.

The model

0.85 AUC

A consent-propensity model on Snowflake that scores about 300K physicians.

Illustrative ranking

Sales-rep view

  1. A physician
  2. Web-enrichment agent
  3. Personalized outreach strategy

Admin view

Snowflake Cortex chatbot

School and toolbox

B.S. Computer Science (Co-op), Toronto Metropolitan University

Graduating
May 2028
Honours
Dean's List 2023 to 2026
Courses
Data Structures and Algorithms, Object-Oriented Programming, Software Engineering, Operating Systems
Certification
AWS Certified Solutions Architect, Associate (SAA-C03). In progress, expected 2026.

Languages

  • Python
  • TypeScript
  • JavaScript
  • SQL
  • C/C++
  • Java

Frameworks

  • React
  • Next.js App Router
  • React Native Expo
  • Node.js
  • FastAPI
  • Tailwind CSS
  • React Query

Backend and data

  • PostgreSQL
  • Supabase
  • pgvector
  • REST APIs
  • ETL pipelines

Cloud and tools

  • AWS
  • Snowflake
  • Docker
  • CI/CD
  • Git/GitHub
  • Power BI

Say hello.

Recruiting, a project, or a question about anything here. Email is the fastest way to reach me.

nawaf.mahmood2005@gmail.com