Jayden Brar · Toronto, Canada

Building intelligence for decisions that matter.

Applied AI, product data science, experimentation, ML engineering, quantitative research, and open-source systems—connected by a single question: how can intelligence make consequential decisions more inspectable?

Explore the fieldwork Trace the research
Applied intelligence / research instrumentScroll to calibrate
Applied AIProduct data scienceExperimentationML engineeringQuantitative researchOpen-source engineeringEmerging researchApplied AIProduct data scienceExperimentationML engineeringQuantitative researchOpen-source engineeringEmerging research

Controlled experiments and public engineering

Work that leaves an evidence trail.

Four systems. Distinct states. Every figure and ownership boundary kept visible.
Language embedding / ML framework experimentComplete

Sentiment extraction, rebuilt for the CPU.

Not everything needs a GPU.

Benchmarked a full embedding × ML framework matrix: 16 language-embedding types across multiple classifier families, including random forest and logistic regression, while re-optimizing spell-checking libraries, multithreading, RAM-aware chunking, and sparse matrices for blazing-fast, lightweight, and lossless social-media analysis.

Undergraduate Program Findings AwardeeDirectly recognized by the Program Director.
16 × MLembedding / framework matrix
O(n)text processing
  • BERT
  • Random forest
  • Logistic regression
  • scikit-learn
  • Polars
  • PySpark
  • PyTorch
Applied software / financial systemsComplete

AML Fraud Detection Deep Learning Model

Multi-model fraud-detection ensemble using an autoencoder, ReLU and sigmoid activations, Adam optimization, and anonymized financial-transaction data.

99.9993%reported ensemble accuracy
  • Python
  • TensorFlow
  • Keras
  • scikit-learn
  • Adam
MoonshotAI / kimi-cliPR #2114
As seen on
CodexClaude Code

Granular permission-policy architecture

Designed and implemented command, MCP-tool, and directory policy with glob matching, protected workspace boundaries, persistent approval state, and regression coverage.

Commits
9
Docs
EN / 中文
Stack
Python · TOML · pytest · MCP
Inspect the open pull request
Google Gemini CLIFormal P1 triage

From unresolved issues to assigned ownership.

Synthesized related unresolved CLI issues into a coherent problem statement, designed the remediation architecture, advanced the work through maintainer review to formal P1 triage, and was assigned ownership of the fix.

Related reports remain scattered

Move through the field to attract the reports, constraints, and maintainer context into one problem.

One line of inquiry, widening in scope

From human aid to simulated decision systems.

EducationSept. 2021–Dec. 2025

B.Comm. Honours, Business Technology Management

Toronto Metropolitan University

Data Science & Visualization specialization, substantial computer-science coursework, Dean’s List standing, and top-3% academic performance.

$7,000BMO Capital Markets Scholarship
Past affiliationSept. 2021–May 2023

Queen’s University Conflict & Analytics Lab

Queen’s University × Oxford researchers

Co-developed a lifesaving legal-tech platform connecting thousands of North American users with local aid, delivered early machine-learning features, and coordinated multidisciplinary delivery across engineers and lawyers.

Incoming · proposed researchSept. 2026–Apr. 2028

Thesis-based M.Sc.

Deep Reinforcement Learning for Land-Use Optimization & Property-Price Modelling

Planned MicroModel research system connecting simulated property-market states, admissible land-use actions, policy learning, and inspectable decision traces.

$30,000Graduate Fellowship and Research Awards
Current affiliationJun. 2026–Aug. 2028

Laboratory of Innovations in Transportation (LiTrans)

Toronto Metropolitan University

Connecting transportation systems, machine learning, simulation, land-use optimization, and property-price modelling.

Official affiliation

Illustrative interaction · proposed research · not empirical results

Watch context sharpen a property-price forecast.

Toggle the signals a model can see. The network gains context while its estimate closes in on a fixed benchmark.
Predict PropertyFive signals1 / 5 connected
Illustrative model alignment
Model forecast
$480,000
Benchmark value$800,000
Forecast accuracy
60.00%
Forecast $480,000 at 60.00% accuracy.
Working paper · in progressNot published · not peer reviewed

Deep Reinforcement Learning in Property Markets via MicroModel Simulation.

A planned manuscript about simulated property markets, deep reinforcement-learning policies, land-use decisions, and interpretable decision traces.

Planned manuscript field

The research framing connects policy learning with an inspectable simulation layer: not only what action a model selects, but how the modelled environment changes and how that trajectory can be examined.

Public dialogue / distinct participation states

Technical work carried into the room.

Upcoming participation as a panelist in September 2026.

Moderated a past Wealthsimple Foundation discussion with PennyDrops TMU.

Built

Business Analysts Student Association

Founder & President · Toronto Metropolitan University · Oct. 2023–Apr. 2024

Led

Wealthsimple Foundation @ Pennydrops TMU

VP of Corporate Relations · Sept. 2024–Jul. 2025

Won

1st

out of85+ Contestants

Tri-Annual Intern Innovation Challenge, where participants take an idea from prompt to presentation before Digital Product Team leadership. Jayden was the first participant ever to receive a special invitation to go beyond the challenge format and present to the Digital Innovation Executive Leadership Team.

Recognized

Nearly $100k in Scholarships

Awards from Companies, Universities, and Research Institutes

Evidence is part of the interface.

Follow the work beyond the portfolio.

Open pull request

Permission architecture, inspected in public.

MoonshotAI kimi-cli pull request #2114 contains 9 commits, automated regression coverage, and English/Chinese documentation for command, MCP-tool, and directory permission policy.

Inspect pull request #2114