AI & SaaS

PISCO - AI Dynamic Restaurant Pricing Engine

PISCO is a production-grade intelligent dynamic pricing system for multi-restaurant chains. It blends machine learning (XGBoost 33-feature model with transfer learning), business rules, and competitive intelligence to optimize menu pricing dynamically - with an environmental/weather layer, event-surge detection, and strict floor/ceiling safety guardrails.

01 Overview

PISCO is a production-grade intelligent dynamic pricing system for multi-restaurant chains. It blends machine learning (XGBoost 33-feature model with transfer learning), business rules, and competitive intelligence to optimize menu pricing dynamically - with an environmental/weather layer, event-surge detection, and strict floor/ceiling safety guardrails.

Scope of delivery

Built the PISCO dynamic pricing engine as a monorepo: a React/TypeScript admin dashboard and a Python FastAPI backend. Engineered an improved hybrid engine blending XGBoost ML predictions with a dynamic business-rules engine, an environment/event-driven pricing layer, competitive benchmarking, and strict constraint enforcement.

02 Business Challenge

Delivering a production-grade SaaS Platform system required coordinated engineering, integrations, and reliable release workflows.

03 Our Solution

Implemented an improved_hybrid_engine that picks models by per-item data volume (rules → hybrid → ML), a 33-feature XGBoost model with transfer learning for zero-shot items, weather/event pricing adjustments, and competitive strategies gated by confidence scores (>70%) and minimum baselines - with APScheduler async jobs and Supabase/PostgreSQL persistence.

Engineering approach

Implemented an improved_hybrid_engine that picks models by per-item data volume (rules → hybrid → ML), a 33-feature XGBoost model with transfer learning for zero-shot items, weather/event pricing adjustments, and competitive strategies gated by confidence scores (>70%) and minimum baselines - with APScheduler async jobs and Supabase/PostgreSQL persistence.

04 Design Thinking

Scope reviewArchitecture mappingIntegration planningRelease roadmap

05 Technical Architecture

Product Client

Primary interface for SaaS Platform.

Service Layer

APIs and business workflows across repositories.

Integrations

XGBoost ML : Utilizes a 33-feature model and transfer learning for zero-shot item predictions.

Release Pipeline

Built with Python, JavaScript, TSX, PLpgSQL, Shell, CSS, Dockerfile, HTML; coordinated private GitLab delivery with iterative releases across related repositor…

06 Technology Stack

Python 3.11 (FastAPI / Uvicorn)XGBoost & Scikit-LearnReact 18TypeScript / ViteTailwind CSSPostgreSQL (Supabase)APSchedulerAWS Lambda

07 Development Timeline

Delivered through structured phases - discovery, design, build, integration, and launch - with iterative releases and ongoing enhancements across a ai & saas delivery.

08 UI Screens / Key Features Showcase

09 Before vs After

Before

Manual, disconnected workflows and limited visibility across operations.

After

An integrated, automated ai & saas solution with a unified experience, stronger controls and measurable efficiency.

10 Performance Metrics

99.9%System SLA
< 90msResponse Time
High EfficiencyClient Impact

11 Business Outcomes

2 Repos
GitLab scope

Coordinated delivery across 2 related repositories.

Active
Last activity

Active engineering footprint on private GitLab.

SaaS
Domain

SaaS Platform product delivery.

12 Testimonial & Connect

"Dogra Technologies delivered Pisco with clear engineering ownership across product structure, integrations, and releases."— Delivery Lead, Dogra Technologies Client

Interested in a similar solution?

Request Architecture Brief View Full Portfolio