Harshil Koladiya

Developed and deployed 'Transfer Truth,' a live LLM data pipeline achieving ≥90% news-source reliability, by implementing an automated 15-case evaluation harness that proactively blocked regressions before release.

Automated comprehensive job-market analysis of ~3,300 roles per run, reducing processing time to under 11 minutes by integrating 4 REST APIs with LLM scoring and a 3-model fallback 'Minions' pipeline.

Contributed to an open-source Node.js AI-agent system by designing and merging a writing-quality guardrail across 4 instruction modes, receiving maintainer review for an 8-file scoped differential.

Optimized AI-agent telemetry, reducing latency from 9.3s to 2.3s and correcting a false error rate of 41.7% through end-to-end debugging.

Increased new-payee straight-through-processing (STP) from 47% to 71% by analyzing 40-60 daily operational interactions using SQL and Excel, identifying friction patterns, and implementing formal process improvements.

Reduced first-contact escalations by 15-20% through root-cause analysis of workflow failures, aligning front-end user actions with backend attributes and compliance flags.

Facilitated SEPA scheme implementation by collecting regulatory and compliance requirements from stakeholders and developing comprehensive configuration and case-note documentation.

Ensured 90-95% QA accuracy across customer profiles and transaction histories by rigorously testing rule-aligned decisioning and validating case data using SQL and Excel.

Processed high-volume daily transactions and meticulously reconciled pricing and SKU discrepancies, ensuring strict compliance with pharmacy regulations and operational procedures.

Increased session completion by 10% through in-depth analysis of funnel data using dashboards and SQL, identifying and resolving critical drop-off points in core product flows.

Enhanced task completion by approximately 18% by leading user interviews, redesigning key screens based on feedback, and conducting iterative user testing.

Managed the product backlog and orchestrated Scrum rituals, including writing detailed user stories and facilitating effective sprint reviews and stakeholder check-ins.

Achieved significant user growth, expanding daily active users to 200K, and improved click-through rates by 25% through optimized A/B-tested paid campaigns across Meta and Google platforms.