Dakshin Ramesh

[email protected] • +965 60330745 • Peelamedu, Coimbatore District-Tamil Nadu, India

PROFESSIONAL SUMMARY

Highly motivated Software Developer with expertise in Deep Learning (RoBERTa, LSTM) and Natural Language Processing (NLP), specializing in complex Sentiment Analysis and Ensemble Modeling. Proficient in Python, Java, and TensorFlow/Keras, with hands-on experience in corporate IT infrastructure, including network and cloud security. Proven ability to translate complex data challenges into robust, user-centric solutions.

WORK EXPERIENCE

•Information Technology Intern
05/2025 - 06/2025
Naser M. Al-Baddah & Partner Gen. Trdg. & Cont. Co. W.L.L. (NBTC Group), Kuwait , Kuwait
• May 2025 – June 2025 (1 month)
• Completed an intensive internship in the Information Technology Department, gaining direct exposure to enterprise IT infrastructure and systems.
• Assisted in foundational IT operations, including the configuration of network systems to ensure connectivity and performance across the organization.
• Participated in projects focused on deploying cloud computing solutions, enhancing the company's scalability and modernizing data storage capabilities.
• Supported the IT security team in implementing security systems, demonstrating a strong commitment to data integrity and system protection.
• Showcased a strong interest and dedication to understanding and improving various aspects of the corporate IT environment.

EDUCATION

Bachelor of Technology
01/2022 - 01/2026
PSG College of Technology, Coimbatore , Coimbatore
Higher Secondary Certificate(HSC)
01/2022
DPS FAIPS,KUWAIT , Kuwait GPA: 70%
Secondary School Certificate(HSC)
01/2022
DPS FAIPS,KUWAIT , Kuwait GPA: 93%

SKILLS

Technical Skills: Deep Learning , Sentiment Analysis , Emotion Classification , Network System Configuration , Cloud Computing Solutions , Security System Implementation , Natural Language Processing (NLP) , Ensemble Modeling , Decision Fusion , Hyperparameter Tuning.
Soft Skills: Technical Problem Solving , Data Analysis , Collaborative Teamwork , Research & Literature Review , Vendor Coordination , Client Servicing , Time Management (Project Timeline)
Tools: Deep Learning: TensorFlow/Keras , PyTorch , Transformers (RoBERTa). Languages: Python. UI/Deployment: Gradio. Data & ML: scikit-learn , Pandas , NumPy.
Other: RoBERTa (Transformer Architecture) , LSTM (Recurrent Neural Networks) , Class-Weighted Loss , Real-Time Inference , Project Documentation (Final Review).

PROJECTS

Stock Trading Strategy using Reinforcement Learning
Technologies: Python, PyTorch, yFinance, NumPy
• Developed a REINFORCE policy gradient agent to automate trading decisions using historical stock data
• Built a simulated trading environment with discrete Buy/Hold/Sell actions for training the agent
• Designed and trained a deep neural network-based policy network to maximize cumulative rewards
• Integrated real-time stock data using yfinance and implemented reward normalization for stable learning
• Trained the agent over 1000 episodes, resulting in improved decision-making and profit optimization
Text Mining and Sentiment Analysis using Deep Learning
Technologies: Deep Learning Frameworks: TensorFlow/Keras, PyTorch, Transformers. Models: RoBERTa (Transformer), LSTM (Recurrent Neural Network). Tools & Libraries: Python 3.9+, Gradio (for Web UI), scikit-learn, Pandas, NumPy
• This project developed a hybrid Deep Learning ensemble model for robust, multi-class Emotion Classification from text data (joy, anger, sadness, fear, love, and surprise). The core innovation is a hybrid approach combining the contextual Transformer-RoBERTa with a sequential LSTM model. The outputs of these models are synthesized using a Weighted Decision Fusion Module, where weights are dynamically calculated based on each model's F1-Macro scores to prioritize the most reliable prediction. The final system is deployed as a user-friendly, high-precision, real-time analysis tool using a Gradio interface.

CERTIFICATIONS

Internship Training in Information Technology
06/2025
Naser M. Al-Baddah & Partner Gen. Trdg. & Cont. Co. W.L.L. (NBTC Group)

ACHIEVEMENTS

• Achievements
• Engineered a Hybrid Deep Learning Ensemble Model for multi-class emotion classification (joy, anger, sadness, fear, love, surprise).
• Developed a Weighted Decision Fusion Module that dynamically calculated weights based on F1-Macro scores to combine predictions from RoBERTa and LSTM, maximizing the reliability and accuracy of the final system.
• Improved Model Robustness by integrating the contextual power of RoBERTa with the sequential learning capabilities of LSTM, overcoming the limitations of relying on a single model for complex emotional data.
• Mitigated Data Imbalance using a Class-Weighted Loss approach, ensuring fair prediction performance across rare emotions and validating success using the Macro F1 metric.
• Delivered an End-to-End Solution by implementing the final fused model using Python, TensorFlow/Keras, and the Transformers library, and deploying it with a user-friendly, real-time Gradio interface.

Similar Resumes

MUHAMMAD ASIF High School Teacher House #38 Cnanal Breez Ali Town Raiwind Road Lahore, Punjab, Pakistan ASHOK Yennam, Venkata Krishna Reddy Project Manager Aachen,DE JASMEET ARORA SEO Intern Noida ANYI LISETH HOYOS TAPIERO Bogotá D.C. · Colombia