About Skills Projects Experience Contact
Available for opportunities

Hi, I'm Ricky Ansari

|

Building intelligent systems at the intersection of machine learning, quantitative finance, and software engineering.

View My Work Get In Touch

About Me

I'm a software engineer and quantitative developer passionate about building systems at the frontier of technology. With deep expertise in quantitative finance, machine learning, and full-stack engineering, I architect solutions that bridge rigorous mathematics with production-grade software.

My work spans from ultra-low latency trading engines leveraging NVIDIA B200 GPUs to serverless AI pipelines on AWS — with a strong mathematical foundation in stochastic calculus, causal inference, and deep learning. I'm driven by solving problems where performance, correctness, and scale all matter.

When I'm not coding, you'll find me exploring cutting-edge research papers, contributing to open-source projects, or diving deep into the latest advances in AI and quantitative methods.

5+ Years Experience
20+ Projects Built
15+ Languages & Frameworks
Ricky Ansari

Skills & Technologies

Languages

Python TypeScript C++ Rust Go Java Julia Mojo OCaml R Q/KDB+ SQL CUDA

ML & AI

PyTorch JAX/XLA TensorFlow scikit-learn XGBoost Hugging Face torchsde signatory LightGBM SHAP

Quantitative Finance

Neural SDEs Path Signatures Conformal Prediction Causal Inference Monte Carlo VaR/CVaR Heston Model HMM GARCH Black-Scholes

Frontend

React Next.js WebGPU D3.js Tailwind CSS deck.gl Zustand

Backend & Cloud

AWS Node.js FastAPI PostgreSQL MongoDB Redis GraphQL Docker Kubernetes

DevOps & Tools

Git GitHub Actions Terraform Prometheus Grafana Jupyter Vercel

Featured Projects

Featured Project

Elite Quant System

Production-grade multi-language AI trading platform featuring Signature-Informed Transformers, Temporal Fusion Transformers, Deep RL (PPO/SAC), Conformal Prediction, and multi-agent LLM reasoning. Implements the complete technology stack used by elite quantitative firms including Citadel, Jane Street, and Two Sigma.

Python PyTorch C++/CUDA Rust Go KDB+ Julia OCaml
View on GitHub
Featured Project

Project Nanotron

Single-node NVIDIA B200 quantitative engine optimized for ultra-low latency trading. Features Monte Carlo Tree Search with prior networks, GPUDirect Storage for NVMe-to-GPU data paths, formally verified C++23 execution layer, and WebGPU L3 order book visualization at 120Hz. Sub-10µs signal latency.

Mojo JAX/XLA C++23 Rust KDB+/q WebGPU React
View on GitHub

Neural SDE Portfolio Optimizer

Mathematically rigorous portfolio optimization combining Neural Stochastic Differential Equations, Path Signature methods from Rough Path Theory, and Conformal Prediction for distribution-free uncertainty quantification. Features Wasserstein Distributionally Robust Optimization.

Python PyTorch torchsde signatory CVXPY
GitHub

Causal Discovery Trading Engine

Trading system that goes beyond correlation to discover causal relationships using PCMCI+, Attention-based Causal Transformers, and Counterfactual Reasoning. Features causally-constrained attention masking and interventional predictions for alpha generation.

Python PyTorch tigramite NetworkX
GitHub

TransformerQuant

Transformer-based quantitative trading system applying multi-head self-attention mechanisms to financial market prediction. Features temporal pattern recognition across multiple asset classes with positional encoding adapted for irregular time series.

Python PyTorch Transformers Pandas
GitHub

BXMA Risk Quant Platform

Risk management and quantitative analytics platform prototype with production-grade VaR/CVaR calculations, stress testing scenarios, sensitivity analysis, and real-time portfolio risk monitoring across equity, credit, and derivatives positions.

Python NumPy SciPy Pandas
GitHub

AWS Document Processing Pipeline

Production-grade serverless document processing pipeline using AWS Step Functions, Textract, and Bedrock Batch Inference with Knowledge Bases. 13-step async architecture with WaitForTaskToken pattern and Distributed Map for massively parallel document extraction.

TypeScript AWS CDK Step Functions Bedrock Lambda
GitHub

VIX Strategy Engine

Volatility-based trading strategy exploiting VIX term structure, volatility risk premium, and mean-reversion properties. Features robust walk-forward backtesting, regime-aware position sizing, and dynamic hedging with VIX futures and options.

Python Pandas SciPy Matplotlib
GitHub

Heston Stochastic Volatility Model

Implementation of the Heston stochastic volatility model for option pricing with Monte Carlo simulation, characteristic function methods (Fourier inversion), and calibration to market implied volatility surfaces using Levenberg-Marquardt optimization.

Python NumPy SciPy Matplotlib
GitHub

SDE Numerical Methods

Comprehensive library for solving Stochastic Differential Equations using Euler-Maruyama, Milstein, and higher-order Runge-Kutta schemes. Includes strong/weak convergence analysis, stability testing, and applications to mathematical finance and option pricing.

Python NumPy SciPy Matplotlib
GitHub

Boomerang Physics Simulator

Computational physics simulation modeling boomerang flight dynamics including aerodynamic lift and drag forces, gyroscopic precession, and turbulence effects. Features 3D trajectory visualization, parameter sensitivity analysis, and wind condition modeling.

Python NumPy SciPy Matplotlib
GitHub

LISFLOOD Hydrological Model

Modified LISFLOOD rainfall-runoff hydrological model validated against Hurricane Harvey flooding data. Enhanced soil infiltration calculations achieving ~15% accuracy improvement and ~10% runtime reduction. Integrates QGIS geospatial analysis and NOAA precipitation data.

Python QGIS NetCDF NumPy NOAA Data
GitHub
View All Projects on GitHub

Experience

Full-Stack Software Engineer

Rainmaker Market Systems
Sep. 2025 – Jan. 2026
  • Architected the central matching engine that connects LPs, GPs, SPs, and startups by transforming investor questionnaire data into structured profile vectors across 75 criteria, computes pairwise weighted compatibility scores using cosine similarity, and returns ranked results via paginated REST APIs cached with Redis, achieving <50ms P95 latency across 10K+ profiles
  • Led development of the React/Next.js frontend with multi-step user flows and reusable component library; built Node.js/Express backend services with MongoDB aggregation pipelines, compound indexes, and Zod schema validation spanning 300+ data fields

Quantitative Valuation Summer Analyst

Houlihan Lokey
May 2025 – Jul. 2025
  • Performed mark-to-market valuations for private equity and hedge fund portfolios across equity, credit, and derivative instruments using DCF, Monte Carlo simulation, Black-Scholes and binomial lattice option pricing, and credit yield-spread decomposition
  • Built a Snowflake Document AI + Snowpark Python pipeline that ingests raw valuation PDFs and images, runs OCR extraction, maps fields to canonical schema, validates parsed values, and writes completed models to Excel; reduced model-build time by 60%
  • Automated equity risk premium (ERP) estimation for the Technical Standards Committee by building a Python framework that pulls Damodaran implied ERP, Kroll/Ibbotson historical premia, and live S&P 500 data, replacing previous 5-hour manual process

Data Science Associate

AIG (American International Group)
Jul. 2022 – Jul. 2024
  • Developed Python analytics pipelines that ingest claims and premium transaction data from in-house platforms, compute 25+ operational KPIs, fit models to premium-collection curves, estimate claim severity tail parameters, and run Kaplan-Meier survival analysis on open claim resolution times; automated daily refresh with threshold breach alerting
  • Built Power BI and Tableau dashboards serving C-suite, actuarial, and underwriting stakeholders across North America, Europe, and Asia; reduced monthly reporting cycle by 5 business days and eliminated manual Excel consolidation across regional offices
  • Enhanced proprietary data warehouses and analytics platforms by redesigning dimensional models, converting stored procedures, and adding targeted indexes after diagnosing bottlenecks; achieved 40% reduction in query execution time across reporting feeds

Quantitative Trading & Portfolio Management Intern

Compak Asset Management
Jun. – Aug. 2020; Jun. – Aug. 2021
  • Rebuilt tactical asset allocation model in Python by implementing a regime-switching signal ensemble to rebalance across equity, fixed income, and cash; tuned entry and exit thresholds with walk-forward grid search over 500+ parameter combinations
  • Extended technical factor library and evaluated alpha candidates for standalone predictive power via rank information coefficients
  • Designed proprietary “Score Plus” multi-factor scoring model that computed cross-sectional percentile rankings and sorted stocks into quintile buckets; evaluated long-top/short-bottom quintile spread returns and compared metrics to sector benchmarks

Computational Geophysics & Climatology Research Intern

Mandil Group, Columbia University Dept. of Applied Mathematics
Jun. – Aug. 2020; Jun. – Aug. 2021
  • Developed adaptive mesh refinement (AMR) configurations within the GeoClaw package to simulate storm surge inundation; designed AMR hierarchies that dynamically refine grid cells, achieving 3-4x improvements in spatial accuracy and runtime
  • Backtested the LISFLOOD hydrological model against observed flood inundation data from Hurricane Harvey; optimized codebase through addition of early-exit logic and replacement of infiltration loop with vectorized NumPy implementation; reduced simulation runtime by 10% and improved numerical accuracy by 15% by eliminating floating-point truncation error

Get In Touch

I'm always interested in new opportunities and collaborations. Feel free to reach out!

Email f.ansari@columbia.edu
LinkedIn ricky-ansari
GitHub github.com/fansari100
Say Hello