Navneeth Krishna
BLR | SFO | ✞ HOU
Navneeth Krishna is a Web, Android, & Full-Stack Software Engineer, a Machine Learning Researcher, and an Adventurer.
He is a M.S. in Computer Engineering graduate from NYU Tandon School of Engineering and a B.E. in Computer Science & Engineering graduate from DSCE, Bengaluru, India. As a Machine Learning / AI Engineer at Calidris / Radiant Data, Navneeth plays a pivotal role in advancing innovative solutions like AIKKA and designing sophisticated credit risk models. Navneeth is a globe-trotter and enjoys blogging, stargazing, and catching Pokémon on his journey. He actively volunteers and manages leadership roles in several domains in his free time.
Skilled in a number of front- and back-end tools and practices for web development & design.
Proficient in Software Development with Java/Python/C/C++.
Drafts & curates content for social media management and corporate relations.
Develops Android apps using Android Studio with Java/Kotlin.
Researches a span of machine & deep learning methodologies for various domains.
Trained in nature and general photography with a portfolio on Instagram.
Developed and documented a video subtitle generator application to enable file upload of video, generate subtitles based on real-time audio transcription with Deepgram, and burn subtitles into the video. Created a live subtitle generator for a live video stream using OpenAI’s Whisper API and GPT-3.5 for translation. Utilized standard software engineering practices & creativity in documentation.
Implemented a full-stack application built with React.js and Python Flask + MongoDB/PostgreSQL. It is an educational PoC finance app in development that provides various tools & analysis for managing investment portfolios with collaborative effort using HTTP methods like GET, POST, PUT, and DELETE. The app authenticates OpenAI API keys with principles of RESTful APIs.
Developed a YouTube Video Summarizer based on GPT-3.5-turbo and Whisper API by OpenAI. Validates YouTube link and OpenAI API Key, downloads the video, transcribes the video's audio, and summarizes the content. Frontend on HTML, CSS, and JS. Backend uses Firebase NoSQL Cloud Firestore database, Python Flask, Whisper API, gpt-3.5-turbo, pytube, etc. Interaction uses RESTful APIs.
Developed a website to solve the trending ‘Wordle’ game with a team of 2 based on iterative feedback & text analytics. Deployed an AWS EC2 instance to host a front-end built on HTML, CSS, & JS to feed user input for processing. Set up a backend server to service front-end requests and securely transfer data through REST APIs.
Developed a full-stack application with user authentication and CRUD functionalities as the captain of a team. Designed & developed a relational schema following database design conventions for a graduate-level course requirement. Utilized Python Flask for the web server and deployed the front- and back-end on Heroku.
Created a back-end system with Python & Flask to service REST API calls from the front-end and the research hosts. Designed and developed an impactful relational database by leading a team of 3 and meeting 100% research goals. Created a front-end web application using React to service user requests and maintained a robust architecture. Optimized regression models by training & testing research data with 2.5 M entries in order to reduce empirical loss.
Utilized Residual Neural Networks on classifying 220k+ classes of artworks from ‘The Met’ dataset. Developed and trained 10+ varieties of CNN backbones in a parametric approach alongside Contrastive Learning. Derived and documented insights obtained through KNN hyperparameter tuning and performance evaluation metrics.
Utilized Residual Neural Networks to solve the CIFAR-10 Classification problem with 94.3% test accuracy. Developed and trained batches using Stochastic Gradient Descent, CNNs, Skip Connections, and Cross-Entropy Loss. Derived and documented insights and hyperparameter tuning outputs throughout the process.
Developed a model that performs character recognition from images of license plates captured from CCTV footage. Transformed Support Vector Machines and employed Connected Component Analysis. Implemented a Python script that reads image input of license plates, detects characters, and prints the characters.
A restaurant management Android application built with Android Studio and SQLite. Mamma Mia employs a modern-day intuition of managing a restaurant's diners at ease. Features include User Checkin, Table Allocation, Order Placement from Menu, and Bill Calculation.