Kishore Kumar

I hold a Master of Science in Artificial Intelligence and Machine Learning with First Class Distinction from Coimbatore Institute of Technology, Tamil Nadu, India. I have hands-on experience developing machine learning models, analyzing data, and solving complex problems using AI techniques. I am proficient in Python, TensorFlow, Keras, OpenCV, LangChain, LangGraph, and other ML libraries, with knowledge in deep learning, natural language processing, and computer vision. Eager to learn and explore new AI technologies, I seek opportunities to work on challenging projects and contribute to the field of AI and ML.

As someone who was into computers and programming from a young age, the field of AI & ML gives me the platform to explore various applications and push the boundaries of technology. I am fascinated by the potential of artificial intelligence and machine learning to revolutionize industries and improve the quality of life for people around the world. I am excited to contribute to the development of AI and ML technologies that can solve complex problems and have a positive impact on society. Whether it's developing innovative algorithms, creating intelligent systems, or exploring how existing fields can be improved using Machine Learning.

Highlights:

CryptoBERT, a BERT model fine-tuned on custom Crypto currency related sentiment dataset, that has garnered over 2.7 Million lifetime downloads
Organized the Open Expo at Melina 2023, providing a platform for students to display their work, featuring over 30 innovative projects.
Played the integral role in development of Multi-Agent system which powered Verk by SuperAGI
Was responsible for development and maintenance different automated AI employees at Verk, creating a flawless end-to-end automated pipeline
Graduated with First Class Distinction, achieving a CGPA of 9.2

I am on the job market for 2024.

Experience

SuperAGI
AI Engineer Intern
January 2024 - Present
Banglore, India
- Played a vital role in development of Multi-Agent System architecture with Agent-to-Agent communication protocol
- Published the extensive research on Multi-Agent System as a detailed blog. Multi-Agent System
- Developed a library of tools that can be used by the agents to solve a user specific task by intelligently communicating among each other
- Designed the asynchronous subprocess communication among agents, which reduced the response time by 65%
- Worked on automatic code development tool based on SWE-Agent, improving it by adding summary based file indexing and core execution loop
Bosch Global Software
Project Trainee
July - December 2022
Coimbatore, India
- Created a centralized database and developed an analytics dashboard to view the relationship between data by integrating data from diverse data sources, resulting in 30% reduction in data retrieval
- Developed an Incremental Machine Learning tool with 93% accuracy for analyzing and identifying errors present in log files and to automatically identify and learn new errors that occur over time
- Explored Zero-Shot Learning, Incremental Learning, Topic Modelling and PreTrained NLP algorithms and compared their performances to choose the best baseline, which improved the existing accuracy from 3-5% depending on the model
SeaportAI
AI and DS Intern
Oct - Nov 2021
Chennai, India
- Enhanced detection rates by 15% through the implementation of automated machine learning models using VertexAI Workbench.
- Performed a comprehensive evaluation of various outlier detection algorithms and packages to determine their effectiveness and appropriateness for the project
- Developed and implemented a robust testing script to systematically evaluate and compare the performance of different outlier detection models.

Research Highlights

Personality Prediction Based on Twitter Tweets
Kishore Kumar R,Jeevasurya V,Shana J
CVR 2021. (Part of the Algorithms for Intelligent Systems book series)
paper code bib
Enhancing Intrusion Detection Systems: Unleashing the Power of Hybrid Optimization Approaches
Jeevasurya V,Kishore Kumar R,Sreekanth S,Hemashree P
ICASIT 2023. (Proceedings Indexed by Scopus and Crossref)
paper code

Projects

Resume-Job Matching Tool
- Developed an API tool which matches the user's resume with the job description and gives the matching score
- Uses LangChain, Vector Database (ChromaDB), and LLM (OpenAI GPT 3.5)
- Extracts the required and relevant content from a resume and matches with job description
code
ASL Sign Prediction
- Built a CNN model which predicts the shown ASL (American Sign Language) sign
- Used the Sign Language MNIST dataset to train and evaluate the model to get 95% accuracy
- Made a utility program with OpenCV to capture the user's sign and show the predicted output
code
Crypto Market Sentiment
- An API program, that takes in keywords from the request and finds its sentiment on Google News and Twitter tweet feed data.
- Uses Google API (GNews) to extract data from Google news and Twitter API(snscrape) to scrape Tweets
- Sentiment of the data is analyzed using FineTuned transformer model, CryptoBERT
code
CNN For Sentiment Analysis
- Applied and measured CNN's performance on Textual data by using Glove Embedding for creating word vectors
- Conv1D and Maxpooling1D are used to extract the features and build the model, which produced 83% accuracy
code