poyboi

• Deployed a Q&A platform for government clerks and bureaucrats to quickly access domain-specific information, reducing dependency on manual escalation while enabling over 5,000+ users to resolve queries autonomously • Engineered modular RAG pipelines using LangChain, Unsloth, FastAPI, & PostgreSQL, with fine-tuned LLMs and document pre-processing to support scalable inference across government domains (like elections and taxation) • Handled full-stack responsibilities including Chainlit UI, PM2 process management, data ingestion from 20K+ PDFs and CSVs, and thorough documentation of all ML/data workflows.

• Developed an AI-powered cold email generator that scrapes a LinkedIn profile to create personalized marketing emails, tailored to user behavior and industry vertical, saving companies 60% in marketing expenses • Designed a waterfall scraping system using LinkedIn API, ScrapingDog, and Lix, paired with LangGraph templates and custom prompts, enabling personalization across 1,000+ context-aware emails weekly.

• Built an interactive Web3 NFT experience with customizable assets and cloud-based AI-driven user engagement • Leveraged scalable REST APIs for authentication and LLM inference using Modal’s GPU-backed infrastructure.