I build AI systems that turn complex data into high-stakes business decisions.
Data science leader focused on risk, capital allocation, fraud, and decision intelligence. My work connects models to financial outcomes, operational action, and enterprise-scale execution.
$100M+
Capital Allocation Influenced
$10M+
Projected Annual Savings
87%
Fraud Alert Reduction
👉 Building systems where models don’t just predict → they decide and create VALUE
How I think
From models to decisions
Most organization do not fail at AI because they lack models. They fail because predictions never become operational decisions. My work is built around closing that gap – translating analytical signal into action that leaders can trust. measure, and scale.
- Data → signal → decision → action → outcome
- Model performance matters only when decision quality improves
- Risk, governance, and execution are part of the product
- AI should reduce noise, increase confidence, and accelerate action
From data to business outcome
1
Data
Operational, portfolio, fraud, and market inputs
2
Models
Risk scoring, forecasting, classification, optimization
3
Decision Engine
Prioritization, policy action, allocation guidance
4
Impact
Lower loss, better efficiency, stronger ROI
Risk
Cost
Growth
How I Do
From prediction to financial impact
I turn models into decisions and decisions into measurable business impact.
This is the system behind that transformation.
Data Sources
Portfolio, operational, market, and fraud signals
Feature Engineering
Structured representations for decision quality
ML Models
Risk scores, forecasts, anomaly detection
Decision Engine
Rules, prioritization, and next-best-action logic
Business Actions
Intervention, policy change, allocation, escalation
Financial Impact
Loss reduction, savings, efficiency, ROI
Selected Impact
Business outcomes, not just models
These are representative examples of how I structure data science work: start with a business decision, build the analytical system around it, and tie the result to measurable financial or operational value.
5% → <2%
Portfolio Risk Intelligence
Resident risk classification and early intervention
- Reduced portfolio delinquency for more than 5% to less than 2%
- Enabled proactive intervention instead of reactive collections
- Unlocked more than $10M in projected annual savings
87% ↓
Fraud Detection Optimization
Signal prioritization with model-driven risk scoring
- Cut suspicious fraud alert volumes by roughly 87%
- Improved investigator precision and analyst throughputÂ
- Maintained control discipline in a regulated environment
$100M+
Capital Allocation Decision Support
Predictive intelligence for market and portfolio decisions
- Supported high-stakes investment and expansion decisions
- Improved confidence in market entry and risk-adjusted allocations
- Connected predictive analytics directly to strategic action
Leadership and Scale
Built for enterprise adoption
My role is not limited to building models. I lead teams, align stakeholders, navigate governance, and ensure that data products become real operating capabilities.Â
Team leadership
Leading 5+ data scientists across strategic initiatives
Budget responsibility
Managing more than $2M in AI and analytics work
Stakeholder coverage
Partnering with Finance, Risk, Operations, and senior leadership
Execution focus
Driving deployment, adoption, and measurable business use
Scale and Influence
1
You
2
Team
3
Stakeholders
4
Systems
5
Enterprise Impact
Capabilities
What I build
I focus on analytical systems that improve the quality, speed, and economics of decision-making.
Decision intelligence systems
Predictive risk modeling
Fraud and anomaly detection
Capital allocation analytics
LLM-powered analytics copilots
Time-series forecasting
Model governance and deployment
Executive-facing KPI systems
Enterprise AI & Data Science Strategy
Quantitative Risk Modeling
Credit Risk Optimization
Team Building & Cross-Functional Leadership
Core Stack
Python
SQL
CatBoost
AWS
XGBoost
Snowflake
Power BI
Tableau
Pandas
Matplotlib
I design for both analytical rigor and organizational usability – the model, the workflow, the reporting layer, and the decision context all need to work together.
Quantitative Research and Trading
A second edge: systematic thinking in markets
Alongside enterprise AI leadership, I explore systematic trading and quantitative research using signal discovery, validation, and risk-aware portfolio design. This broadens the story from enterprise analytics to generalizable decision systems.
CAGR
25-35%*
SHARPE
Research-led
FOCUS
Risk-adjusted
- Regime-based strategy design across equities and ETFs
- ML-driven signal discovery using tools such as CatBoost and nearest-neighbor approaches
- Walk-forward testing, Monte Carlo resampling, and drawdown-aware positioning sizing
Career Highlights
Data Science Leader specializes in driving enterprise AI/ML strategy, quantitative risk analytics, and decision intelligence across financial services, real estate, and automotive sectors.Â
Known for aligning advanced analytics with business outcomes.
Financial Services & Banking
Real Estate & PropTech
Core Skills
Enterprise AI & Data Science Strategy
Quantitative Risk Modeling & Fraud Analytics
ML Model Development, Deployment & MLOps
Portfolio & Credit Risk Optimization
Predictive & Prescriptive Decision Intelligence
Model Governance (MRM / MDC) & Regulatory Compliance
Data Product Development & KPI Frameworks
Capital Allocation & Financial Performance Analytics
Team Building & Cross-Functional Leadership
Automotive & Manufacturing
Enterprise Analytics & Digital Transformation
Task/Activity:
Key Actions Taken:
- Architected and launched a high-precision Resident Classification Model to identify delinquency risk early.
- Deployed predictive default and churn models to enable proactive intervention and extend resident lifetime value.
- Advised Finance and Operations executives using ML-driven portfolio risk intelligence to guide strategic portfolio decisions.
- Built and led a team of five data scientists, establishing strict model acceptance criteria and scalable ML governance frameworks to accelerate delivery and adoption across departments.
Case Study
AI-Driven Portfolio Risk Reduction & Capital Strategy
Situation:
At Progress Residential, the company managed a large single-family rental portfolio with increasing delinquency exposure and inefficient capital allocation signals. Finance and Operations leaders needed stronger predictive intelligence to guide $100M+ capital decisions and proactively reduce portfolio risk.
Objective:
Lead the enterprise AI strategy to reduce delinquency risk, improve capital allocation precision, and operationalize predictive intelligence across the organization.
Results:
- Reduced delinquency rate from 5%+ to below 2%
- Unlocked $10M+ in projected annual savings
- Influenced $100M+ in capital allocation decisions
- Improved decision speed and executive adoption of AI-driven strategy
- Strengthened cross-functional trust in predictive analytics and enterprise AI initiatives
Task/Activity:
Key Actions Taken:
- Engineered ML-driven momentum and regime-based volatility strategies, validated through robust cross-validation and walk-forward testing.
- Built end-to-end Python research pipelines to automate data ingestion, feature engineering, signal generation, and backtesting, accelerating alpha discovery.
- Implemented disciplined portfolio construction, incorporating volatility targeting and strict drawdown controls to optimize capital efficiency.
Operationalized systematic trading workflows to enable continuous strategy iteration, signal validation, and scalable portfolio management.
Case Study
ML-Driven Trading Strategy & Risk-Disciplined Portfolio Management
Situation:
After founding Shreyan Trading LLC, proprietary capital was deployed across equities, ETFs, derivatives, and crypto during highly volatile market regimes. The objective was to generate institutional-grade risk-adjusted returns while maintaining strict downside discipline in non-linear markets.
Objective:
Design and operate a systematic trading framework capable of identifying momentum and volatility regime shifts while protecting capital through disciplined risk controls.
Results:
- Delivered 35–45% annualized returns
- Achieved 2–3× performance vs. the SPY benchmark during the same period
- Maintained maximum drawdown limited to 18%
- Achieved a Sharpe ratio of 1.83
- Compounded returns through disciplined reinvestment and risk management
Task/Activity:
Key Actions Taken:
- Engineered an XGBoost-based fraud risk scoring model layered on top of the existing rule framework.
- Replaced binary rule-trigger alerts with a probability-based risk ranking system, allowing only high-confidence cases to proceed to manual investigation.
- Partnered with compliance and model risk governance teams to ensure audit transparency, regulatory alignment, and model validation readiness.
- Integrated the scoring model into the fraud detection workflow to prioritize investigations and optimize analyst capacity.
Case Study
AI-Driven Fraud Detection Optimization
Situation:
 At USAA, credit and debit card fraud detection relied primarily on rule-based triggers. Any breached rule generated an alert, yet only 2–3% of alerts represented legitimate fraud. Each alert required manual review, creating significant operational waste, investigator fatigue, and regulatory pressure around efficiency and control effectiveness.
Objective:
 Reduce false positives while maintaining full regulatory compliance and ensuring 100% capture of true fraud cases.
Results:
- Reduced suspicious alert volume by 87%
- Maintained capture of all true positive fraud cases
- Significantly improved investigative precision and prioritization
- Reclaimed substantial analyst hours and increased case throughput
- Sustained full regulatory compliance while modernizing fraud detection architecture
Task/Activity:
Key Actions Taken:
- Designed and implemented an automated KPI validation pipeline to replace manual SQL and Excel workflows.
- Built a system that ingested business-generated KPI input files and reconstructed KPI logic programmatically.
- Implemented structured SQL transformations to automatically recalculate metrics and verify report accuracy.
- Developed automated comparison and variance detection, flagging discrepancies with detailed audit-ready reporting.
- Generated a final validation output formatted for compliance documentation, ensuring transparency for regulatory review.
Â
- Partnered with risk and compliance stakeholders to align the automation with governance standards and position it as a control-strengthening initiative.
Case Study
Automating Enterprise KPI Validation & Governance
Situation:
Within USAA’s Enterprise Risk organization, analysts were responsible for validating high-impact enterprise KPI reports generated by business units. Validation required analysts to rebuild reports manually using SQL queries, Excel pivot tables, and complex reconciliations, creating a process that was time-intensive, repetitive, and vulnerable to human error while remaining critical for governance and regulatory transparency.
Objective:
Redesign the validation workflow to automate KPI verification, eliminate manual reconciliation effort, and maintain strict governance and audit defensibility while improving team productivity.
Results:
- Reduced work effort by ~720 hours per analyst annually
Â
- Significantly lowered manual reconciliation risk
Â
- Improved validation accuracy and consistency across KPI reports
Â
- Accelerated KPI certification and governance timelines
Â
- Recognized by leadership and expanded as a model for additional automation initiatives
Task/Activity:
Key Actions Taken:
- Developed machine learning forecasting models using historical sales data, seasonality patterns, regional demand signals, and market indicators.
- Engineered time-series forecasting frameworks to predict dealership-level demand with greater granularity.
- Partnered with supply chain and production planning teams to integrate predictive forecasts into manufacturing decision cycles.
- Implemented model monitoring and recalibration pipelines to adapt predictions as market conditions and demand signals shifted.
Â
- Transitioned the organization from reactive production adjustments to proactive, demand-driven manufacturing optimization.
Case Study
ML-Driven Demand Forecasting & Supply Chain Optimization
Situation:
At Nissan North America, supply chain volatility and inconsistent dealership demand forecasting created production lags, inventory imbalances, and rising handling costs. Overproduction in some regions and understocking in others disrupted manufacturing flow and impacted market stability. Leadership required more accurate demand forecasting to better align dealership inventory with production planning.
Objective:
Improve sales forecasting accuracy and build predictive models to optimize supply-demand alignment, stabilize production schedules, and reduce excess inventory costs.
Results:
- Reduced inventory handling costs by 12%
- Minimized production lags and stabilized manufacturing cycles
- Improved supply-demand alignment across dealerships
- Enhanced operational predictability and market responsiveness
Testimonials
What People Say
“I had the opportunity to work closely with Visharath Adhikari on building an ML-powered fraud detection system, and his leadership and technical depth truly stand out.
Visharath led a cross-functional team of data scientists, data engineers, and business stakeholders to develop an XGBoost-based risk-scoring model that reduced suspicious alert volumes by 87%, a transformative impact on operational efficiency and precision.
What impressed me most was his ability to deliver under resource constraints. He brings clarity to complexity, aligns technical strategy with business outcomes, and drives execution with discipline and vision.
Visharath is a rare combination of strong AI capability, practical data science leadership, and decisive execution. Any organization would benefit from his expertise.“
“I’ve had the pleasure of working with Visharath, and I can confidently say he is an exceptional data professional. He has a strong ability to translate complex business problems into well-structured analytical approaches and build robust models that drive real impact.
What sets him apart is not just his skill in testing and comparing multiple modeling techniques, but his discipline in selecting the right approach based on business context and performance trade-offs. He doesn’t stop at model development — he ensures end-to-end implementation by building reliable pipelines that move model outputs seamlessly into business workflows.
His work consistently combines technical depth with practical execution, making his solutions both sophisticated and usable. Any team would greatly benefit from his analytical rigor, problem-solving mindset, and ability to turn data into actionable results.“
Connect
Let's build decision intelligence systems that scale
I’m interested in opportunities where AI drives measurable business impact – especially across risk, strategy, capital allocation, and quantitative systems.