TL;DR

Nova started as a borderline automation tool used for scraping LinkedIn and automating outreach. Faced with increasing legal and ethical concerns, the platform needed to pivot—fast. The new vision: help sales teams and recruiters enrich and keep CRM/ATS profiles up to date using LLM-powered agents and semantic search.

My Role

As Lead Architect and Full-Stack Developer, I guided the MVP team’s transition to a modern stack before eventually rebuilding the platform solo for the v1 launch. I integrated LLMs at a time when enterprise-grade AI tooling was still in its infancy and took ownership of every layer of the product—architecture, backend, frontend, AI integration, and product direction.

Tech Stack

To bring Nova to life, I chose a stack that balanced speed of development with long-term scalability. On the frontend, Next.js gave me the flexibility to move quickly while still delivering a polished user experience. The backend ran on .NET Core with GraphQL APIs, which made it easier to serve structured data consistently across different integrations. For search and data handling, I paired Elasticsearch with ArangoDB, which allowed me to model complex relationships between people, companies, and roles. Security was a priority from day one, so I implemented modern authentication protocols like OAuth with Auth.js. Finally, I leaned on the early versions of OpenAI APIs to build semantic enrichment agents, turning raw profile data into actionable intelligence long before such tooling became mainstream.

Challenges & Breakthroughs

Nova faced a critical challenge in pivoting from a legal gray area to a fully legitimate AI-driven SaaS platform. Achieving this required delivering a production-grade LLM integration in just a few weeks, surpassing the performance of a large data science team that had spent months iterating on rule-based models. At the same time, the full system architecture was carefully designed to ensure scalability, modularity, and compatibility with multiple CRMs and ATSs, making the platform robust, flexible, and future-ready.

Key Features Built

CRM/ATS Integration

I built Nova to live inside the tools recruiters and salespeople already used. Instead of asking them to switch platforms, I engineered deep CRM and ATS integrations that synced data in real time, prevented duplicates, and made profile enrichment feel invisible. It turned rigid, legacy systems into adaptive, AI-driven environments, all while keeping workflows intact.

AI Enrichment Pipeline

Scaling Nova meant tackling massive data. I architected an enrichment pipeline capable of processing millions of profiles, cleaning, deduplicating, and enriching them with AI. It transformed noisy, fragmented datasets into structured intelligence that recruiters and sales teams could actually trust—an engine powerful enough to keep pace with the ever-changing global talent pool.

Persona/Job Matching

Traditional keyword matching wasn’t enough. I developed AI models that analyzed skills, experiences, and inferred traits to connect candidates with jobs or buyer personas more intelligently. This went beyond surface-level filters—surfacing hidden talent, uncovering relevant prospects, and giving teams a sharper lens to find the right fit faster, with a precision legacy tools simply couldn’t match.

Chrome Extension

To bring Nova’s intelligence closer to users, I designed a Chrome extension that blended seamlessly into ATS and CRM pages. It detected missing information, flagged outdated records, and suggested AI-driven updates on the spot. With real-time notifications and one-click actions, users could instantly enrich and sync data, turning tedious data upkeep into a natural part of browsing.

Outcome

Nova is now used by leading recruiting and sales firms in Canada, empowering them to work with enriched, actionable, and compliant data—without leaving their current tools.

Download Nova Sneak Peek (PDF - 2MB)

I’m especially proud that I was able to deliver in just a few weeks what an entire data science team had struggled to achieve in over a year. The result not only outperformed their models but also offered a better user experience and full LLM integration, proving the impact of a lean, focused approach to product development.