Faizan Khan
Software Engineer → Quantitative Finance
Software engineer who gets pulled toward hard problems, where the model, the system, and the data pipeline all have to work together. Started in full-stack, drifted into ML engineering, currently pointed at quantitative finance.
Background
About Me
The Honest Version
I'm a software engineer who gets pulled toward hard problems, the kind where the model, the system, and the data pipeline all have to work together. I started in full-stack and drifted toward ML engineering. That means when I'm working on a problem, I'm thinking about the pipeline, the API, what happens when the model drifts, and how you actually retrain it. Not just getting the accuracy number up in a notebook.
I ended up pointed at quantitative finance because it's one of the few places where both sides matter equally. You need to understand the math well enough to trust your signals, and you need to engineer well enough that the system actually runs in production.
I'm not trying to do everything. I want to find the problem where the math and the engineering both have to be right, and spend a long time getting good at that specific thing.
Profile
Point of View
"Most quant researchers can't ship a production system. Most engineers don't really understand what the model is doing. I'm trying to close that gap. The three projects on this site are where I'm testing whether I actually can."
Background
Education & Experience
Education
CUNY Brooklyn College
Brooklyn, NY
- Relevant coursework: Data Structures, Analysis of Algorithms, Data Tools & Algorithms, Machine Learning
- Minor in Data Science
Experience
Brooklyn College
Brooklyn, NY
- Engineered a production ML pipeline for NERIS (a federal firefighter reporting system) using the Anthropic Claude API and fine-tuned models to extract structured data from unstructured narratives at scale
- Built Python ETL pipelines with automated validation, schema enforcement, and anomaly detection, cutting manual processing time by 40% and reducing data quality incidents by ~30%
InZone Inc.
Remote
- Built and shipped a distributed microservices backend in Node.js/Express serving 5,000+ daily active users, owning architecture decisions from design through production
- Containerized multi-service applications with Docker and GCP achieving 99.9% uptime; mentored 4 engineers on API design and coding standards
AutoLake LLC
San Francisco, CA (Remote)
- Drove a 25% increase in system throughput at a B2B data lake infrastructure company by standardizing RESTful API contracts across engineering teams
- Found and fixed 15+ security vulnerabilities (XSS, CSRF) through a production audit against the OWASP top 10
Scaled active membership to 100+ students; coordinated technical workshops on AI engineering, full-stack development, and systems design.
Portfolio
FinTech Projects
Three projects where I tried to build the real thing, not just the notebook version.
Quantitative Trading Framework
I wanted to know if momentum and mean-reversion signals actually work when you test them honestly: real transaction costs, out-of-sample periods, no cherry-picking. This is that experiment.
Credit Risk Scoring Engine
I built this to understand how banks actually score loan applications. Raw loan data, XGBoost model, a deployed REST API, and SHAP explanations for why each decision was made. The whole stack, not just the model.
Real-Time Fraud Detection
Built solo over 8 weeks. Transactions stream through Kafka, features are cached in Redis for 2ms scoring, and XGBoost catches 89% of real fraud while flagging only 0.08% of legitimate transactions. That tradeoff was the whole design challenge.
Capabilities
Technical Skills
Machine Learning
- XGBoost / LightGBM
- SHAP Explainability
- Scikit-learn
- Imbalanced Learning (SMOTE)
- MLflow Experiment Tracking
- Optuna HPO
- Feature Engineering
Data Engineering
- Apache Kafka (KRaft)
- Redis Feature Store
- PostgreSQL / TimescaleDB
- ETL Pipelines
- SQLAlchemy Core
- SQLite Time-Series Store
- Pydantic v2 Validation
FinTech Domain
- Credit Risk & Scorecard Design
- Fraud Detection Systems
- Strategy Backtesting
- Adverse Action Explanations
- Sharpe / Sortino / Calmar
- Transaction Cost Modeling
- Walk-Forward Validation
Cloud & DevOps
- Google Cloud Run
- Docker / Docker Compose
- FastAPI
- GitHub Actions CI/CD
- Streamlit
- Plotly Dash
- Vercel / Render / Netlify
Right Now
Quantitative Trading Framework: walk-forward optimizer is working, but the out-of-sample Sharpe always comes in lower than in-sample. That is probably correct, not a bug.
Advances in Financial Machine Learning by Lopez de Prado. In the combinatorial purged cross-validation chapter right now. Dense. Have read it twice and I think I understand it.
Whether TimescaleDB is good enough for tick data or if kdb+ is worth the learning curve. kdb+ is everywhere in HFT but the syntax is its own thing and the licensing is not cheap. Still deciding.
Let's Connect
Get In Touch
Wrapping up at Brooklyn College in May 2026 and looking for full-time roles in FinTech and quant finance. If you want to talk about any of the work, reach out. And if you want to argue about whether Happy Death Day counts as horror, I'm available for that too.
© 2026 Faizan Khan · Built with Next.js + Tailwind CSS