• Composer • Artistic Director • AI/ML Developer • Quantitative Researcher
.jpg)
ABOUT
Amin Sharifi is a composer, AI and machine learning systems developer, and quantitative researcher whose work bridges mathematical music, advanced quantitative trading systems, and interdisciplinary research at the intersection of art and technology. His career is defined by a fascination with patterns, structures, and complexity, whether in the mathematical frameworks underlying his compositions or in the data-driven models powering advanced financial systems.
MACHINE LEARNING & QUANTITATIVE RESEARCH
Sharifi’s work in machine learning and quantitative finance centers on developing advanced algorithmic trading systems that blend AI-driven techniques with rigorous quantitative analysis. In this domain, he has independently created sophisticated trading frameworks, demonstrating how cutting-edge research can be applied to real-world financial markets. His projects span reinforcement learning agents that adapt to market patterns, complex event processing for multi-timeframe signal detection, and rigorous risk management mechanisms, all reflecting a deep integration of theory and practice.
​
Case Study: QuantRL: Deep Reinforcement Learning for Trading
QuantRL is a modular reinforcement learning framework that couples a custom trading environment with a Proximal Policy Optimization agent using a self-attention policy architecture. The system separates market and portfolio feature spaces, supports both discrete and continuous actions, and encodes practical market frictions in the environment so that training objectives align with risk and execution reality. Reward functions target risk-adjusted utility rather than raw return, and the training stack supports deterministic experiment seeding, checkpointing, and ablation of architectural components to measure causal impact on performance.
Evaluation follows strict time-series protocol. Walk-forward analysis and embargoed cross-validation prevent leakage, Monte Carlo resampling probes path dependence, and out-of-sample tests benchmark against naive and simple rules. Reporting captures a full distributional view of behavior through rolling performance curves, tail risk diagnostics, turnover and capacity estimates, and stability metrics across regimes. The framework is engineered for extensibility so researchers can introduce alternative objectives, hierarchical policies, or ensemble routing while preserving reproducibility and auditability. Github Repo
​
Case Study: AlphaFX: Multi-Timeframe Automated Trading System
AlphaFX is a production-grade, event-driven system that orchestrates signals across multiple time horizons through complex event processing. A regime layer based on Gaussian Mixture Models segments the market into distinct states, and regime-aware feature pipelines apply context-specific thresholds and stabilization logic to reduce false positives. Signal consensus is enforced across timeframes so that short-horizon patterns align with higher-level structure before orders are considered, and risk is governed by volatility-bounded sizing, correlated exposure control, and drawdown brakes that act at both instrument and portfolio scope.​
The platform mirrors live constraints during research by replaying historical event streams through the same orchestration logic. Post-trade analytics quantify risk-adjusted performance, efficiency of signal routing, behavior under shocks, and sensitivity to frictions such as spread, slippage, and latency proxies. Performance reports include equity and drawdown curves, distributional diagnostics for trade returns and durations, and regime-conditioned summaries that reveal how behavior changes across states. The system serves as a template for integrating statistical structure, machine perception, and formal risk controls into a single cohesive workflow. Github Repo
Read More About AI & Quant Projects
​
Music Composition
​​
Sharifi’s music is internationally recognized for its conceptual depth, structural innovation, and visual-musical integration. His works have been performed in the United States, France, Germany, Italy, Greece, Lithuania, Russia, Canada, Poland, the Czech Republic, and Iran. Pianist Kathleen Supové has described his music as “creative, individualistic, artistic,” and the Memphis Daily News has called it “the product of an unbridled imagination.” His compositions have been brought to life by ensembles including JACK Quartet, Mivos Quartet, Ostravska Banda, Hypercube, A&C String Quartet, and the Tuscanini Philharmonic Orchestra, and have appeared at festivals such as ReMusik, Ostrava Days, Synthetis, eviMus, Tehran Contemporary Music Festival, Druskomanija, WSU Contemporary Art Music Festival, and RISUONANZE. Renowned soloists including Christopher Otto, Austin Wulliman, Alex Sopp, Verena Rojc, and Futaba Niekawa have championed his work.
​
Notation as Visual Language
Sharifi’s practice sits at the confluence of sound, mathematics, and visual art. Drawing on a background in computer science and programming, he creates original notation systems that function both as performance scores and as standalone visual works exhibited in the United States and Europe. His research addresses the relationship between composer, performer, and audience, aiming to bridge the composer-listener gap with visual structures that remain accessible while preserving formal complexity. Works such as Maximum Insufficiently Identical Outlines and Minimum Different Enough Details probe perceptual thresholds by crafting materials that are structurally similar yet nearly indistinguishable to the ear. His notational language, rooted in morphology theory, cubist approaches to time, and the concurrency of irrational tempos, gives composers precise control over every musical parameter and opens new avenues for experimentation.
​
Computational Methods in Composition
Many of Sharifi’s pieces integrate computational processes and custom software. In Koocheh Baghi, he analyzed an improvisation on the Persian tar using a Python-based spectral tool of his own design, classifying harmonic spectra into distinct categories and mapping timing with nanosecond precision. The resulting data replaced traditional meter and tempo, producing an organic rhythmic architecture inside intricate contrapuntal textures. In Aposynthesy, a deconstruction of Parviz Meshkatian’s Booy-e-Baran, he quantized every frequency and rhythmic value, then reconstructed the material through dynamic stretching, compression, and timbral transformation. Across Mise-en-scène, Mise-en-synthesis, and Aposynthesy, the evolving notation becomes a compositional language that can be tailored to the needs of each work.
​​
Multimedia and Thematic Projects
Sharifi’s projects often extend beyond the purely sonic into multimedia, conceptual, and socio-environmental realms. Salt Flat Portals combines generative visuals, Afrofuturist performance, documentary footage of Utah’s Great Salt Lake, and chamber music. Distorted Landscapes addresses the destruction of Utah’s natural environments through electroacoustic textures, fractal-generated visuals, and symbolic sonic gestures. His music balances formal rigor and expressive freedom, creating kaleidoscopic experiences in which structure, pattern, and texture are as central as melody or harmony.
​
Mentors and Honors
His teachers include Sven-David Sandström, Nader Mashayekhi, David Dzubay, and Don Freund, and he has studied with Chaya Czernowin, Toshio Hosokawa, Krzysztof Penderecki, Georg Friedrich Haas, Raphaël Cendo, and Oscar Bianchi. In 2017 he won first prize in the XXIII Concorso 2 Agosto International Composition Competition for TrombionOphone or Riders in the Field of Hope. He has served as Assistant Director of the IU New Music Ensemble.
​
Entrepreneurship and Artistic Direction
As founder and artistic director of Petrichor Records, Sharifi supports contemporary classical music by living composers. He has championed new voices and provided release and performance opportunities for young musicians, resulting in more than twenty albums in recent years and extensive collaborations with artists across the globe.
Petrichor Records Official Website
​
Education
-
PhD (ABD) – Music Composition, University of Minnesota (Graduating 2025)
Advisor: Dr. Guerino Mazzola
- Developed machine learning models for audio signal processing.
- Implemented real-time spectral analysis algorithms using Python and Max/MSP.
- Created data visualization tools for complex audio datasets using matplotlib.
- Designed and implemented neural networks for sound synthesis and analysis.
-
MIDS – Master’s in Information & Data Science, University of California, Berkeley (2025–2027)
-
MA – Music Composition, Duke University
- Developed Python-based tools for large-scale musical data analysis.
- Implemented statistical analysis methods using NumPy and pandas.
- Created custom algorithms for pattern recognition in time-series data.
- Applied machine learning techniques to audio classification problems.
-
MM – Music Composition, Indiana University Bloomington
- Developed software for real-time audio processing.
- Implemented mathematical models for sound synthesis using Python and SuperCollider.
-
BM – Tehran University of Art
- Ranked #1 in graduating class.
- Best undergraduate thesis: "Mathematical Structures in Contemporary Music."
- Extensive coursework in algorithms, data structures, and computer programming.
THE ME
I started composing and programming when I was 9.
Before that I had been playing tar, the Persian long-necked lute, and I was already in love with patterns. In high school I went deep into mathematics and computer programming, then chose music composition in college because it felt like the most direct way to shape structure and time. The center of my practice has always been the same curiosity about how form creates meaning.
My most formative mentor was Nader Mashayekhi. He took our composition class into Iran’s mountains and deserts, where we made sound installations and conceptual performances in open landscapes. He taught me to close my eyes, listen to space, and hear structure as something physical. Those experiences rewired how I think about notation, timbre, and the connection between an idea and its realization.
​
I write music the way I write software. I design systems, define constraints, and let materials evolve inside them. I have built my own tools for spectral analysis and developed notational languages that read as visual art and operate as precise performance scores. I am drawn to perceptual thresholds, to materials that appear identical yet differ just enough to change how you listen. I like work that is exacting and imaginative at the same time.
​
The same mindset guides my AI and quantitative research. I build learning systems that search for structure in nonstationary data, validate ideas with care, and aim for clarity under risk. Whether I am shaping a quartet, training a policy, or editing a score, I am after the same thing: a clean architecture that reveals pattern, invites interpretation, and holds up under pressure.