shravyamalla

Built an LLM-powered chatbot that serves as a personalized content strategist, helping users create and schedule brand-aligned social media content tailored to growth goals such as increasing follower count, engagement, and click-through rates. Designed and implemented the entire chatbot pipeline using LLMs and agent-based architecture with prompt engineering and fine-tuning models to enable goal-driven, platform-specific responses. Integrated a multi-layered analysis engine into the chatbot to support personalized content generation and strategy across multiple connected accounts using insights from historical posts, brand guides, formats, and performance trends. Centralized strategic planning and automation within a single conversational interface, streamlining content ideation, scheduling, and optimization across platforms.

Led a team of 4 in developing two interactive platforms for dynamic dashboard, plot, and report generation: one based on agentic architecture for natural language querying and another using a static pipeline with predefined prompts and libraries. Achieved a 30x reduction in report generation time and a 90% increase in accuracy through R&D and execution. Enhanced system performance and user experience by re-engineering the backend to align with client branding. Helped client acquisition by showcasing product prototypes, and tailored the product for successful integration into existing workflows. Architected a hyper-personalized marketing messenger by breaking a user persona down into components. Built the Proof of Concept for Google using the GCP suite and brought in multiple clients by showcasing the prototype.

Developed upskilling programs in ML and Computer Vision. Performed sentiment analysis on user data gathered by web scraping. Improved inventory management through advanced SQL queries. Created weekly cohort charts to track usage of the app by the users.

Used OpenCV and DepthAI with an OAK1 camera to create a model to track people and assign unique ids to them in python. Modified a cumulative object counting model to track people, achieving a result that improved upon the existing model by 200%. Created a Flask-based API and SQLite database that allowed the user to query directly from a web frontend.