Projects

Redesigned the model from scratch. Created a new dataset combining historical workers’ comp premiums, wages, class codes, and claims data. Developed claim development factors from 10+ years of historical captive data and applied NCCI’s split point methodology to generate a target using primary and ratable excess losses. Built and refined a frequency-severity LightGBM model in DataRobot. Integrated the model into a custom app for the Pricing Team to quote new clients based on estimated annual claims.

Created a self-service voice call reporting tool using Coefficient and Google Sheets. Built cleaned and aggregated Snowflake layers from raw Amazon Connect call data using DBT and Dagster. Scheduled data syncs to Google Sheets and built a tabular + dashboard view to replace manual reporting. Logged process insights and limitations for future self-service data product development.

Developed a data source in Snowflake using SQL in DBT and Dagster to pull quota, closed-won, and invoice data from Salesforce. Created logic to accurately capture first-invoice annualization for clients with varied pay groups (e.g., weekly, semi-monthly). Leveraged Coefficient to import this data into a dynamic Google Sheet with role-based views (BDM, management). Included an override sheet for manual adjustments where needed.

Self-initiated an end-to-end contact clustering project. Built an aggregated dataset from raw Salesforce data including job role, tenure, communication volume and quality, case metrics (e.g., reopen rate, time to close), sentiment analysis, and NPS history. Ran clustering models in DataRobot and identified 6 distinct contact groups, with strong alignment to NPS segments and referral behavior. Delivered interpretability insights and guidance for future data collection (e.g., contact influence and goals).

Self-initiated a survival analysis project to model client term timelines and break-even points using Kaplan-Meier survival curves. Segmented analysis by service offering, client size, and termination reason (e.g., out of business vs. switching providers). Modeled breakeven for administrative fees versus sales acquisition and servicing costs. Ran Python survival models from a self-service Snowflake data object using dbt and Dagster. Outputs informed retention strategies and product focus.

Proposed a multi-phase project to automate case review using Snowflake Cortex’s SUMMARY function and keyword flagging. Built a prototype report that scored and sorted cases by potential relevance based on flagged terms in email content, subjects, and descriptions. Report included summaries, filters by pay date, and deep links to case details. Future phases included iterative refinement of keywords and training an ML model on labeled case data.

Designed and built multiple self-service data objects covering anticipated and requested data needs for cases, calls, sales, clients, and contact details. Sourced from iSolved HRIS, PrismHR (multiple instances), Salesforce, and claim systems. Defined the database schemas, table structures, and layered data models to support scalable, efficient querying. Created cleaned, typed layers and master/reference tables using DBT and Dagster. Final objects were used for both object-based and snapshot reporting, and loaded into Google Sheets via Coefficient for self-service access.

Used DBT and Dagster to build a Snowflake data pipeline sourced from Salesforce. Created two models in DataRobot using a Random Forest Classifier—one for the first half of the sales process and one for the second half. Ensured proper data partitioning to avoid leakage. Scheduled the model training and scoring to run weekly, storing results in Snowflake (with snapshot history) and loading predictions into Salesforce.

Developed an automated solution to extract encoded employee and client documents via the HRIS API, storing metadata and file references in Snowflake. Built a Python-based decoding pipeline to handle large, encoded data chunks, manage invalid file types, and output files into a structured shared drive using standardized naming and folder conventions. This enabled clean handoff to the new HRIS and allowed document access for clients when needed.

Scraped 6,510 LinkedIn job postings using Python to analyze trends across data job titles. Explored demand by role type, employer industry and size, job location and work mode, seniority, employment type, pay ranges, and most sought-after skills by role. Created visuals and published a two-part Medium article highlighting actionable findings and insights for aspiring data professionals.

Used cleaned email text from Salesforce cases and applied K-means clustering in Python. Focused initially on verb-based features, but resulting clusters reflected actions rather than topic content. Determined that further development was needed to improve semantic relevance, but project was paused due to shift in company priorities.

Built custom Python automation with web scraping using Selenium and HTML parsing to log into iSolved, navigate client profiles, and extract key data unavailable via API. Supplemented scraping with automated report downloads from the Report Center, ingesting CSVs and loading into Snowflake. Enabled clean downstream use for data migration into the new HRIS.

Built a SQL-based text parser in Snowflake to process email content row-by-row. Identified and removed repeated signature blocks, reply history patterns, and automated disclaimers to create a cleaned dataset of email content suitable for text analysis and downstream modeling.

Used the cleaned email dataset to identify common error-related phrases. Created a dedicated table tagging emails based on matching keywords, and built a Tableau report that allowed users to dynamically include/exclude flagged phrases. Enabled fast scanning and review of potential service issue communications.

Developed a join-based approach to calculate business minutes by creating a minute-level reference table with associated lookup values. The solution joins to start and end timestamps, then calculates the difference between lookup values to derive business time. Adopted as the standard method and used for 4+ years.

Built an end-to-end ML pipeline using Salesforce data to predict client satisfaction as Delighted, Satisfied, or Poor. Engineered ~300 features capturing client profile (size, industry, tenure), communication patterns (volume, direct vs. employee, sentiment via Snowflake Cortex, time to resolve), risk indicators (AR balance, layoffs, internal handoffs), product usage, referral history, and summarized NPS trends. Trained a LightGBM model in DataRobot, prioritized top 16 features, handled class imbalance, and automated scoring/prediction flow (Snowflake ↔ DataRobot ↔ Salesforce). Developed feedback loop for future feature enhancement.

Built a company-wide KPI dashboard using raw Salesforce and Amazon Connect data in Snowflake. Created metrics for call responsiveness (answer rate, answer time, voicemails) and case handling (resolution time, reopen rate, case flow). Final dashboard built in Tableau with summary and trend tabs, definitions, and filters. Maintained usage for 1–2 years.

Rebuilt the spreadsheet from scratch, fixing calculation errors and consolidating all pricing and sales model exports into a single file. Added override fields for controlled edits, ensuring clarity and reducing the need for manual formula adjustments. Enabled seamless transitions between sales models with minimal rework.

Independently initiated and built a reporting solution based on the Health Care Cost Institute’s (HCCI) public methodology. Used raw claims data to replicate HCCI metrics and benchmark client-level performance against national data. Created custom reports for large clients showing chronic condition prevalence, utilization rates, prescription share, claim spend, and more. Designed the reports to support future integration with nurse or physician advisory services for tailored wellness recommendations.

Developed an internal website to serve as a comprehensive training and reference tool. It included guides on insurance lines (workers’ comp, health, TDI), forecasting methods, PEO vs. HR-only products, Hawaii’s economic landscape, and internal processes for pricing and insurance renewals. Originally created to train a direct report, it became a key asset still used by Pricing, Risk, and new leadership.

Developed an SSRS report that allowed users to input a client ID and instantly retrieve historical and current rates across all lines of business, along with high-level client context. Designed to be used alongside the client profitability report to support quick, informed repricing decisions.

Built a model using public state unemployment rates and internal profitability data to estimate client shrinkage and margin erosion. Incorporated configurable inputs for temporary rate relief programs. Forecasted changes to overall profitability month by month. Final results were within 5% of actual performance.

Independently created a reporting solution using Microsoft SQL Server and SSRS that integrated billing data from HRP-Prism HRIS with state unemployment insurance (SUTA), workers’ comp (WC) claim data, and overhead costs for insurance and benefits. The report displayed client tenure, industry, service offering, plan details, revenue stream profitability, WC claims history, and worksite employee trends—accompanied by summary graphs for visual context.

Developed custom development factors based on historical claim emergence patterns. Used year-over-year claims data to forecast future claim development at a monthly cadence.

Built a flexible Excel-based model to apply rate modifiers and calculate net workers’ compensation rates. Included margin fields and was used by both underwriters and the internal HR configuration team to standardize rate inputs across processes.

Selected to join a cross-functional team responsible for acquisition due diligence. Forecasted expected profitability from the acquisition using revenue and cost data. Reviewed and validated detailed billing, insurance, and tax records to assess risk and confirm valuation. Collaborated with Finance, Legal, and Executive leadership.

Performed an ad hoc revenue breakdown by source and client segment. Identified that a recent Pricing strategy shift led to significantly lower average bill rates for new clients, dropping the overall average by over 20%. The change had not been forecasted, leading to the financial gap. Provided insights to the CFO and highlighted the need for proactive forecasting tied to future pricing strategy changes.

Prioritized by leadership, led the collaboration with Salesforce consultants to design and test new Salesforce objects. Outlined required fields, data types, workflows, and ensured versioning was implemented for quoting history. Added pricing analytics fields to support future rate benchmarking. Personally tested the build before launch.

Created a comprehensive reporting suite to track and analyze sales performance. Initially developed in Salesforce but transitioned to Tableau for flexibility. Delivered dashboards showing win rates, pipeline stage conversion, leakage, velocity, inflow/outflow, sales funnel progression, loss reasons, market segmentation performance, and sales cycle time. Developed documentation to introduce and contextualize new sales metrics.

Audited actuarial assumptions and validated three reserve models used by external actuaries. Discovered and corrected a $2M miscalculation in the 2018 reserve estimate, influencing financial reporting and planning.

Took initiative to build an internal website that served both teams: for Sales Ops, it included step-by-step submission checklists, annotated examples of good and bad collateral, and visual guides to improve accuracy; for the Pricing Team, it centralized SOPs, reference materials, and documented rare edge-case scenarios to reduce tribal knowledge and ensure consistency.

Managed the end-to-end healthcare renewal process annually, starting in 2017. Pulled data from Microsoft SQL Server and SSRS, merged with carrier and actuarial files in Excel. Modeled financial impacts (total, employer, employee), tested alternative plans, and ensured compliance with renewal fairness regulations and HIPAA. Created a standardized Excel tool reused annually, including high-level summaries. Led internal strategy discussions with the CEO, negotiated with health carriers, and enabled Account Managers with tailored client collateral. Maintained plan structures across tiered, flat, and minimum premium models, with custom quotes handled as needed.

Built quarterly and annual insurance analytics for healthcare, workers’ comp (captive), unemployment insurance, and temporary disability (captive). Pulled data from SQL and SSRS reports, then combined and analyzed in Excel. Iterated directly with the CEO to refine KPIs and visualizations for board presentation.

Analyzed all client data using SQL and Excel to calculate updated rates. Built an Excel macro to auto-generate client-specific collateral sheets showing rate changes. Created a clean dataset for Client Operations and Payroll Configuration to use for contract updates and rate setup. Executed the rollout with zero errors.

Analyzed claim trends across business segments to define and refine standardized rules for quoting health coverage rates. Filed the rules with the Hawaii Insurance Commissioner to ensure regulatory compliance. These rules were implemented by the Pricing Team and guided how actuarial ratings were adjusted for quoting new business.

Used public unemployment fund data, internal claims, and wage history to forecast the state’s upcoming SUI rate schedule and predict client-level renewal rates. Applied the forecast to refine risk stratification strategy. Created forecasts with 100% accuracy at the state level and near-perfect accuracy for client renewals (only a few missed due to rate threshold shifts).

Redesigned the ClientSpace workflow and field structure to require more deliberate review and completion of key steps before advancing. Although it slightly slowed the initial process, it significantly reduced the need for follow-up, improved data quality, and streamlined handoff to Pricing.

Created a centralized competitor repository to track rates, product details, and pricing methods across known competitors. Used the data to generate collateral for the Sales Team to highlight our value proposition during pitches and to guide Pricing decisions when bidding against specific competitors.