Parth – Python, LangChain, D3.js

An AI-powered system designed to automate brand insight generation and audience targeting for digital advertising campaigns. The platform leveraged multi-agent orchestration to analyze brand presence, generate strategic targeting data, and match it with third-party audience segments. Two main parts to the project:

1. Brand Insight Generation: Upon user signup, a multi-agent Brand Insight microservice generated a brand profile including a synopsis, 10 keywords, 10 IAB categories, and 10 personas with demographic data.
2. Audience Matching & Recommendation: Generated insights were matched to third-party audience segments using similarity search, and a composer agent assembled persona demographics into targeted audience recommendations.

Main scope of responsibilities included the following, but were not limited to:

• developed a modular system where specialized AI agents collaborated to automate brand analysis and audience targeting workflows;
• built a Brand Insight microservice triggered on user signup, orchestrating five distinct agents to extract and synthesize relevant data;
• implemented web crawlers and extractors to retrieve and process content from brand websites as input for downstream analysis;
• used Retrieval-Augmented Generation to create accurate and actionable personas aligned with available ad targeting attributes;
• developed a validation agent to ensure the quality of AI-generated insights across four dimensions: structure, cross-referencing, semantic integrity, and business logic;
• mapped generated keywords, personas, and categories to external audience datasets for precise ad targeting;
• broke down complex tasks into subtasks managed by dedicated agents, improving system scalability and maintainability;
• composed demographic segments into simplified, selectable options for end users (e.g., location, device, gender);
• designed a roadmap to fully automate targeting based on campaign goals, using insights, pixel data, and available targeting parameters.

Customer was a digital advertising platform designed to simplify and automate the process of launching multi-channel ad campaigns. Targeted at brands and agencies, it offered a streamlined interface for campaign creation, audience targeting, and performance analytics across platforms like Facebook, Instagram, Google, and programmatic ad networks. The platform emphasized ease of use, leveraging AI to optimize targeting and deliver real-time insights for better ad performance.

Parth's main responsibilities included:

• revamped the analytics engine to support scalable and efficient data processing;
• integrated new data streams to enhance analytics coverage and insights;
• optimized data storage architecture using OpenSearch in combination with AWS Aurora, Redis, and Redshift;
• achieved a 57% reduction in system latency through performance improvements;
• reduced storage costs by 35% by streamlining data retention and storage strategies;
• improved overall operational efficiency and system responsiveness.

This client was an AI-powered advertising automation platform designed to simplify the creation of digital ad campaigns. It enables businesses to input a simple campaign goal and automatically generates optimized, ready-to-launch campaigns by leveraging brand data, user behavior (via pixel tracking), and targeting parameters—streamlining media planning and execution through intelligent agent collaboration.

He was mainly in charge of the following:

• architected the DSP platform and led a development team to deliver key features;
• spearheaded the deployment of the Advanced Campaign Creator;
• developed and implemented reporting dashboards;
• built and managed user management features, including Single Sign-On (SSO);
• designed and integrated various REST APIs to support platform functionality;
• utilized AWS Lambda and API Gateway for serverless backend services;
• implemented authentication and authorization using Auth0;
• developed frontend components using PHP and jQuery plugins;
• managed and optimized multiple data storage systems for performance and scalability.

An asynchronous queuing infrastructure designed to efficiently manage and process large volumes of requests to third-party APIs, ensuring reliable, scalable, and timely integration with external services while minimizing latency and handling rate limits.

Among others, Parth has managed the following:

• led the development and implementation of an asynchronous request queuing system for third-party API integrations;
• analyzed and identified performance bottlenecks causing high user response times;
• designed and implemented optimized code solutions to improve API request handling;
• achieved a 98% reduction in user response time related to third-party API calls;
• collaborated with the team throughout issue discovery, solution design, and deployment phases.

A robust data infrastructure designed to collect, process, and analyze user and campaign data in real-time. The system included a reporting engine for visualizing performance metrics and KPIs, and a Data Management Platform to aggregate first- and third-party data, enabling advanced audience segmentation and targeted campaign optimization. Integrated with DSP and CRM tools, it empowered marketers with actionable insights and improved ROI tracking across digital channels.

Among other's, Parth contributed the following:

• led data architecture design for the reporting analytics engine and Data Management Platform (DMP), ensuring scalability and robustness using modern ETL pipelines with Redshift, Python, PHP, AWS Lambda, and MySQL, reducing report generation time by 65%;
• integrated The Trade Desk API for Real-Time Bidding (RTB) via AWS API Gateway, enabling self-service ad campaign management and ad inventory purchase, cutting support team dependency by 90%;
• collaborated cross-functionally to develop data visualization tools using Bootstrap and jQuery plugins, boosting campaign optimization efficiency by 30% and reducing time-to-insight by 41%;
• built a custom report delivery system using PHP plugins, Redis, MySQL, and CRON jobs, enabling instant access to campaign data, which contributed to a 64% Month-over-Month retention rate;
• improved platform performance by offloading data retrieval to REST APIs during page loads, achieving a 70% reduction in load time and increased user engagement.

An intelligent ad creation tool designed to streamline the campaign setup process. It automatically generates ad creatives and copy based on product data and user goals, using AI-powered product recommendations and targeting insights. Ideal for e-commerce and retail advertisers aiming for fast, data-driven campaign launches with minimal manual input.

Parth's contributions involved the following:

• reduced creative team workload by 63% by implementing tools that automated parts of the content generation process;
• developed and contributed to a web crawler (spider bot) using Python libraries such as BeautifulSoup and Pandas, with data storage in MySQL, to index and store licensed website content for content marketing use;
• designed and integrated the Shutterstock API into the platform, enabling seamless content discovery, selection, and licensing. This enhancement reduced user abandonment by 33% and significantly improved user engagement and satisfaction.

A software system that enables advertisers and agencies to programmatically buy digital advertising inventory across various online channels.

• Was actively involved in the development and maintenance of DSP SaaS platform;
• Developed and led a team of software engineers to develop and add new features in the platform;
• Contributed and monitored Ad delivery network and custom pixel data management.