Lokesh Talamala | Portfolio

About Me

I am a passionate computer science student with a strong foundation in machine learning and artificial intelligence. Currently pursuing Masters at IIT Madras, I combine academic excellence with practical experience in developing innovative solutions.

I thrive in environments that challenge my problem-solving abilities and push the boundaries of innovation. My approach to technology is grounded in curiosity, continuous learning, and a strong analytical mindset. Whether it's exploring theoretical foundations or applying them to solve real-world problems, I enjoy the process of transforming complex ideas into practical outcomes.

Known for my consistency, resilience, and focus, I adapt quickly to new tools and concepts, and I'm always eager to engage in interdisciplinary conversations that blend software engineering with intelligent systems. I take pride in balancing academic excellence with hands-on exposure, and I believe this blend makes me a well-rounded candidate ready to contribute meaningfully in cutting-edge tech environments.

Projects

M.Tech Project

Topic: Artificial Intelligence integration with Wazuh SIEMs

Wazuh, an open-source Security Information and Event Management (SIEM) and Extended Detection and Response (XDR) platform, integrates Artificial Intelligence (AI) to enhance its threat detection, analysis, and response capabilities. This integration primarily focuses on leveraging Large Language Models (LLMs) and Machine Learning (ML) to improve security operations.

AI Internship Project

Topic: LLM-Based Banking Query Assistant using Model Context Protocol

Designed and deployed an AI assistant using the Model Context Protocol (MCP) to enable GPT-4-based querying over customer transaction data stored in PostgreSQL. Implemented a full server–client architecture, migrated it to AWS EC2 for remote access, and evaluated coding assistants like Cursor and Copilot for enhanced productivity.

Natural Language Processing (3)

Sentimental analysis of restaurant reviews
Detecting stress based on social interactions
Google search engine from scratch

Pattern Recognition & Machine Learning (4)

Polynomial Regression with Regularization
Multi-Class Classification with k-NN and Bayesian Models
Neural Network Regression and Classification Modeling
Support Vector Machine Classification and PCA

Deep Learning (3)

Feedforward Neural Network using Backpropagation
CNN Training from Scratch
RNN-Based Transliteration System

Graphics Processing Unit

Topic: CUDA program to find the most cost effective pathways

Developed a CUDA-based solution to compute cost-effective pathways in large graphs using parallel processing. Optimized memory and thread usage for efficient pathfinding, achieving significant speed-up over CPU methods.