AI/ML Intern — PGAGI

Overview

At PGAGI, I worked on production-grade AI systems across automation, conversational AI, generative media, and EdTech platforms. My work focused on designing scalable LLM pipelines, grounded retrieval systems, optimizing latency and cost efficiency, and collaborating with architects and clients to deliver enterprise AI solutions.

Key Contributions

• • Built a Figma plugin generation pipeline that converts design inputs into MJML-based email templates, integrating Asana MCP server and REST APIs while enforcing structured outputs and schema validation.
• • Developed an AI-powered EdTech learning platform for US clients featuring Duolingo-style adaptive learning with grounded RAG pipelines using hybrid retrieval and reranking supported by curated evaluation datasets.
• • Contributed to Cracked-AI, building generative pipelines that produce AI-driven reels through image insertion, video generation, and music synchronization workflows.
• • Engineered Twilio conversational AI systems enabling automated call transcription, summarization, and operational dashboards with request tracing and token usage monitoring.
• • Reduced LLM operational cost by ~35% and improved inference latency through prompt optimization, context window control, caching strategies, and cost-per-request tracking.
• • Collaborated with AI architects to design production AI architectures and technical PRDs, defining system pipelines, deployment strategies, CI/CD workflows, and monitoring approaches for enterprise clients.
• • Designed modular AI plugin systems with resilience mechanisms including fallback handling, retry logic, prompt/config versioning, and reproducible pipelines.

Tech Stack & Tools