Pulsar | AI-powered procurement optimization platform

Pulsar is our own AI-powered procurement optimization platform. It analyzes supplier invoices at scale, finds cheaper alternatives using semantic AI search, and generates savings reports — turning a process that takes procurement teams weeks into minutes. Built in 4 months by a five-person team, Pulsar handles up to 500,000 products in a single analysis run.

Procurement teams at mid-size and large companies spend hundreds of hours every month doing one thing: comparing prices. They receive invoices from current suppliers, then manually search through price lists from alternative vendors, trying to find the same or equivalent products at lower cost. With 100,000+ products across dozens of suppliers, it's a nightmare of spreadsheets, guesswork, and missed savings.

Manual analysis of 100,000 products takes roughly 100 hours — about $4,000 in labor costs. Each invoice takes 15 minutes to cross-reference. And by the time the analysis is done, the recommendations are already outdated. Companies needed a tool that could do this instantly, accurately, and at any scale.

We built Pulsar from the ground up as a full-stack AI platform. Users upload supplier price lists — the system indexes every product, generates AI embeddings, and stores them in a vector database. When an invoice comes in, Pulsar reads each line item, searches semantically across all indexed suppliers, and finds the closest alternatives. Not just by name — by meaning. "Atlantic Cod Fillet 40cs" matches "Cod Fish Fillets, case of 40" because the AI understands they're the same product.

But finding matches is only half the job. Pulsar's AI then evaluates each alternative — comparing units of measurement, package sizes, and prices — calculates exact savings, and generates a clean report highlighting the best option for every product. Procurement managers get a ready-to-act report instead of a mountain of data.

Companies onboard their suppliers into Pulsar and import price lists in any format — CSV, Excel, or PDF. The system parses every line item, normalizes units of measurement (converting between kilograms, pounds, cases, packs), generates semantic embeddings for each product, and indexes everything in a vector database. When new price lists arrive, the system updates the index automatically, keeping recommendations fresh.

Traditional procurement tools match by SKU or exact product name. Pulsar matches by meaning. "Wagyu Chuck Roll 18cs" finds "Premium Wagyu Beef Chuck, 18-case pack" even if the vendor uses completely different naming conventions. The vector search returns ranked candidates with similarity scores, then an AI filter evaluates each match — checking if the products are truly interchangeable, not just similarly named.

Finding alternatives is useless if you can't compare prices accurately. Pulsar handles the hardest part: unit conversion. It compares products across different package sizes and measurement units — calculating how many packages of Alternative B you need to replace Product A, what the total cost would be, and exactly how much you save. Every comparison is apples-to-apples, even when the suppliers use completely different formats.

Every procurement team has rules: "Never substitute branded products with generics," "Only match organic with organic," "Prioritize local suppliers." Pulsar's Instructions feature lets users define these rules per project. The AI incorporates them into its matching logic — so recommendations aren't just mathematically optimal, they're contextually correct for that specific business.

The output isn't a raw data dump — it's a decision-ready report. For each product, Pulsar shows the current cost, all viable alternatives ranked by savings, match confidence, and a recommended "best pick." Procurement managers review exceptions, approve the best options, and export. What used to take a team a week now takes one person an afternoon.

Four months from first line of code to production. Pulsar processes each invoice in under 1 minute (vs. 15 minutes manually), handles up to 500,000 products in a single analysis run, and generates structured savings reports that procurement teams can act on immediately. The 9-stage AI pipeline delivers match accuracy that manual analysis simply can't replicate — because it understands product semantics, not just keywords.

Built as ApolloRise's own product, Pulsar demonstrates our ability to take a complex AI challenge — semantic search, vector databases, multi-stage filtering, unit normalization — and ship it as a polished, enterprise-ready SaaS platform. From architecture to deployment, every layer was built by our team.

Laravel 11, React 18, TypeScript, Redux Toolkit, PostgreSQL, Redis, Qdrant (Vector DB), OpenAI & Anthropic APIs, Python FastAPI, Docker, Nginx, Minio (S3), Laravel Horizon