Hey, I'm Akash Ravandhu

Working with Captain (Y Combinator F25) to build DeepQuery, an open source retrieval benchmarking platform with SOTA RAG implementations and evaluation pipelines quantifying accuracy, speed, and cost metrics across diverse data domains.

Led a team of 4 developers and 1 designer to build a computer-vision-powered microparticle identification and measurement tool for Akina Inc.

• Developed an agentic multimodal Retrieval-Augmented Generation (RAG) system to assist engineers with Nokia Cloud product deployments and troubleshooting.
• Built a Flask-based API and web UI with text streaming, markdown rendering, and a database management dashboard.
• Deployed the service for using Docker and OCP / Kubernetes, utilizing local LLM inference for data privacy.
• Demonstrated Nokia AI-RAN with local LLM inference on far-edge cloud, enabling GPU-as-a-Service business models for telecom operators.

• Led a team of 10 developers to design, develop, and deliver an end-to-end food ordering system for Share Food Share Love Food Pantry, serving over 4,900 individuals in western Cook County of Brookfield Illinois.
• Held weekly stand ups and sprint planning sessions with the team and monthly check-ins with SFSL’s director for feedback.
• Deployed the MERN stack dashboard and SMS service with AWS EC2, serving order requests from 60+ families per week.

• Fine-tuned NER models for structured manufacturing data extraction in an Agile Scrum team for Knudsen Institute.
• Developed 10+ web scraping solutions with Selenium and BeautifulSoup, automating data extraction from over 7,000 links.
• Improved NER model performance through data visualization, pruning, and building an active learning pipeline.

• Developed and deployed an AI-powered UI prototyping system for rapid web design ideation with Hawken AI clients.
• Achieved 4x faster design deployments by distributing frontend build jobs across worker nodes using BullMQ and Redis.
• Reduced deployment failures through automated error-handling, re-generation logic, and build cleanup pipelines.

• Trained, optimized, and compared 4 statistical machine learning models to predict diagnoses of thyroid disease.
• Performed exploratory data analysis, visualization, and preprocessing on thyroid datasets with Pandas and Matplotlib.
• Achieved an F1 score of 0.928 with a Random Forest classifier by performing hyperparameter tuning with GridSearchCV.