Alva Barakhia
Quantitative Intelligence
for Global Commodity Markets
We extract
structural yield.
Alva Barakhia is a proprietary quantitative research firm and principal market participant.
We engineer high-fidelity statistical models and maintain sovereign-grade alternative data infrastructure to isolate and extract structural yield from global commodity networks. We deploy internal capital to systematically assess mathematical convergence and structural correlations across oil, gas, precious metals, and agricultural supply chains.
We operate exclusively on a principal basis and do not solicit external capital, provide advisory services, or act in a brokerage capacity.
02 / Execution
Not alpha.
Operational floor.
Our research horizons span weeks to quarters, but the positions that express those views are large, concentrated, and held across volatile instruments. Managing that risk in real time — across correlated commodity complexes, during liquidity events, through exchange microstructure changes — requires execution and monitoring infrastructure matched to the speed of the venues themselves.
We maintain FPGA-based pre-trade risk systems, direct exchange connectivity, and co-located monitoring at principal commodity venues. This infrastructure is not an alpha source. It is the operational floor beneath a research-driven mandate:
- —deterministic risk controls.
- —deterministic exit capability.
- —deterministic behavior under stress.
We invest in this capability because the cost of a mismanaged position in these markets is asymmetric, and because our fiduciary duty to our own capital requires it.
03 / What We Do
03.1 / Commodity Market Intelligence
Subclause AWhere things
actually are.
Global physical commodity markets produce an enormous volume of unstructured, fragmented, and poorly catalogued data. Shipping manifests contradict port authority records. Regulatory filings lag physical reality by weeks. Satellite imagery tells one story while broker consensus tells another.
We reconcile these discrepancies at scale. Our pipelines ingest, clean, normalize, and cross-reference data from over 200 sources spanning six continents. The output is a statistically rigorous picture of physical commodity flows — where things actually are, not where spreadsheets say they should be.
03.2 / Statistical Modeling
Subclause BMathematics are standard.
Execution is absolute.
Our core modeling stack is built on methods that have survived peer review, stress testing, and live deployment.
Hierarchical models with dynamic prior updating for demand forecasting across fragmented markets.
Regime detection for identifying structural shifts in supply chain behavior before they surface in price.
Dependency modeling for capturing non-linear relationships between correlated commodity pairs under varying market conditions.
Non-parametric estimation for tail-risk quantification in markets where Gaussian assumptions have historically failed.
Decomposition with adaptive signal extraction separating persistent trends from transient noise in physical flow data.
Simulation frameworks calibrated against proprietary datasets for scenario analysis across multi-leg commodity transactions.
We do not fetishize theoretical elegance. The mathematics are standard; the execution is absolute. Our edge is not in methodological novelty, but in the sheer computational dominance with which we deploy it.
03.3 / Alternative Data Infrastructure
Subclause CWe create
comprehension.
Data is only as useful as the infrastructure that processes it. We maintain:
Real-time pipelines handling structured and unstructured feeds from satellite imagery, AIS vessel tracking, port authorities, customs databases, and regulatory bodies across 40+ jurisdictions.
Proprietary architecture processing petabyte-scale data daily with automated quality scoring, deduplication, and anomaly flagging.
Cross-referencing engines that reconcile conflicting data sources and assign probabilistic confidence scores to competing narratives about physical commodity movements.
Cold and hot tiers optimized for both historical backtesting and real-time inference workloads.
Raw data is not our output. We produce comprehension — cleaned, reconciled, statistically validated intelligence derived from physical market observation. The distinction matters.
04 / Our Approach
Stationarity
is a fantasy.
Most commodity models assume stable relationships between variables. We do not. Our systems continuously re-estimate parameters, re-weight factors, and re-calibrate priors. A model that was right in January may be wrong by March. We assume it is, and rebuild accordingly.
Data beats
narrative.
Commodity markets are dominated by storytelling — geopolitical narratives, analyst opinions, conference-circuit consensus. We measure what is measurable and discard what is not. When satellite imagery of grain storage contradicts a bullish analyst note, we go with the satellites.
Discipline over
novelty.
We do not chase exotic models for intellectual vanity. We apply universally proven statistical mechanics with ruthless, industrial-scale precision.
05 / Markets We Cover
Oil & Gas
Upstream production analytics, refinery throughput modeling, storage capacity estimation, and trade flow reconciliation across major producing and consuming regions.
Particular depth in Middle East, West Africa, and Gulf of Mexico supply chains.
Precious Metals
Quantitative tracking of physical gold, silver, and platinum group metals flows through refining, vaulting, and cross-border settlement networks.
Statistical reconciliation of reported versus observed physical movements.
Agricultural Commodities
Yield forecasting, storage estimation, and export flow modeling for grains, oilseeds, and soft commodities.
Integration of satellite-derived crop health indices with port loading data and customs filings.
Physical Trade Finance
Statistical risk assessment for structured commodity trade finance transactions.
Probability-weighted exposure modeling for documentary credit, pre-export finance, and inventory-backed lending facilities ranging from $50M to $1.5B.
06 / Infrastructure
Built for two things:
throughput and reliability.
Our computational infrastructure is engineered in-house across four layers.
Distributed GPU clusters (NVIDIA H100, 8-per-node, multi-region) for model training and inference. FPGA co-processors for latency-sensitive data processing pipelines.
Direct feeds from major market data vendors, satellite imagery providers, and vessel tracking networks. Dedicated 400GbE fiber between processing nodes.
Tiered NVMe architecture with Intel Optane for hot-path data and high-density arrays for historical datasets spanning 15+ years of physical commodity flow records.
99.99% uptime across all production systems. Geographically distributed redundancy. No single points of failure.
We built every layer of this stack in-house. Off-the-shelf solutions do not meet our throughput or reliability requirements.
07 / Mandates & Counterparties
07.1 / Principal Deployment
Subclause APrincipal.
Not intermediary.
The vast majority of our computational infrastructure is dedicated exclusively to our internal balance sheet. We operate as a principal market participant, deploying our statistical architecture to capture structural yield and execute proprietary mandates across global commodity networks. We do not solicit outside capital to fund these operations.
07.2 / Institutional Engagements
Subclause BInfrastructure
exceeds liquidity.
Our infrastructure processes global flows at a scale that vastly exceeds our internal liquidity constraints. Consequently, we selectively accept bespoke mandates from a strictly closed roster of institutional counterparties.
Sovereign entities & state-owned enterprises — confidential, statistically validated market intelligence on physical flows affecting national commodity portfolios and strategic reserves.
Tier-1 trade finance desks — independent probabilistic risk assessments for complex, multi-leg commodity-backed credit exposures and billion-dollar syndicated lending facilities.
Physical commodity majors — quantitative verification for positioning, hedging, and cross-border supply chain reconciliation where traditional broker consensus fails.
Special situations & restructuring — evaluating opaque commodity-linked assets for private equity and distressed debt operators requiring hyper-accurate, satellite-verified physical asset tracking.
- —We do not publish retail research.
- —We do not maintain public data portals.
- —We do not participate in competitive vendor bidding.
All external engagements are initiated by referral, rigorously confidential, and structurally detached from our principal market operations.
08 / Governance & Architecture
Compartmentalized.
By construction.
The firm's principals direct proprietary operations at the intersection of quantitative modeling, sovereign-grade data infrastructure, and cross-border structured finance.
Execution is strictly compartmentalized. We operate a closed partnership model, prioritizing computational density and absolute mathematical discipline over organizational scale.
09 / Contact
All inquiries reviewed
directly by firm leadership.