Aleksandar Perisic

Developed an AI-driven video analytics platform designed to unlock actionable insights from sales calls and customer support video recordings. The goal was to move beyond simple transcription and provide "emotional intelligence" metrics to leadership.

• Multi-modal Sentiment Analysis: Integrated OpenAI Whisper for high-fidelity transcription and custom NLP models to detect shifts in speaker tone and sentiment (e.g., frustration, hesitation, enthusiasm).
• Automated QA Scoring: Built a pipeline that automatically scores video interactions based on key performance metrics, allowing managers to identify "at-risk" accounts without watching hours of footage.
• Searchable Video Knowledge Base: Implemented vector search (embeddings), enabling users to query the video library semantically (e.g., "Show me clips where the client mentioned budget concerns").

This tool transforms raw video data into a strategic asset, reducing churn by alerting teams to negative sentiment trends early in the customer lifecycle.

I created the custom theme for EasyBilanz using Advanced Custom Fields, post types, and features such as a PDF generator with auto-calculated prices, easy filtering with date range search, and the possibility of private invoices. The theme is created following the responsive web design approach.

I designed and built a custom "middleware" application to solve a critical data fragmentation issue for a marketing organization. The client needed to synchronize thousands of records between external proprietary databases and HubSpot without overwriting sensitive manual data entries.

• Selective Bi-directional Sync: Engineered logic to merge data intelligently—preserving manual sales team updates while enriching missing fields from the external database.
• Approval Workflow: Created a React-based "Staging Dashboard" where managers can review, flag, or approve bulk data changes before they hit the production CRM, preventing data corruption.
• Automated Field Mapping: Reduced manual data entry by around 70% by automating the translation of legacy database fields into standardized HubSpot properties.

I built a custom WordPress plugin extending Gravity Forms Survey to turn standard surveys into a guided, high-converting qualification flow. I reworked the UI into mobile-first “button” inputs, added progressive disclosure (“Tell us more” that reveals follow-up fields), and redesigned the submit interaction to reduce drop-offs. I also instrumented the flow for analytics-ready event tracking so marketing teams can see where users abandon and what answers correlate with qualified leads. I can optionally push responses into CRM properties and lifecycle stages for automated follow-up (e.g., HubSpot mapping and dedupe rules).
​

CLIENT VALUE
• Higher completion rate through better UX patterns (button inputs and progressive disclosure).
• Cleaner, more structured lead data (reduces manual cleanup).
• Easier campaign iteration (forms become a “funnel,” not a static questionnaire).

WORK DONE
• Code optimization: Minified and combined stylesheets and scripts to reduce load times and improve render blocking.
• Legacy Upgrades: Refactored outdated codebases to align with modern web standards.
• Cross-browser testing: Utilized BrowserStack to perform real-device testing across iOS, Android, Windows, and macOS platforms for seamless cross-browser compatibility.
• Web performance enhancements: Streamlined assets through image compression and lazy loading. Implemented advanced speed optimization strategies to ensure top-tier Web Vitals scores.
• Caching and CDN integration: Deployed and fine-tuned Cloudflare, CDN, and both server-side and client-side caching mechanisms to boost performance and reduce latency.

These audits significantly improved load times, Core Web Vitals metrics, and overall site efficiency, ensuring an optimal user experience and enhanced technical SEO performance.