Thani Karthik

AI Engineer

https://www.opentalent.in/thani-karthik

AI Engineer with 1+ years in NLP, Full Stack Development & Distributed Systems

Premade Innovations Pvt Ltd

Key Strengths

Strong foundational knowledge in AI/ML, demonstrated through academic projects and internships.
Proficiency in deep learning frameworks (PyTorch, Transformers) and model architectures (EfficientNet, DistilBERT, Llama 3).
Experience with MLOps and deployment tools (Streamlit, FastAPI, Docker) for real-time inference.
Solid understanding of data preprocessing techniques (stratified sampling, handling class imbalance) and optimization (AdamW, BCEWithLogitsLoss).
Practical experience in RAG systems, multilingual NLP, and computer vision applications.

AI is analyzing your overall score…

Identifying your key strengths…

Evaluating your skill match against the job requirements…

Assessing your cultural and operational fit

About

Currently pursuing a Bachelor's in Computer Science and Engineering, I possess 1.2 years of experience in AI/ML and full-stack development. My expertise spans building sophisticated AI systems like multilingual RAG chatbots and digital image forensic classifiers, alongside developing robust web applications with FastAPI and modern frontend technologies. I am proficient in Python, PyTorch, and various data engineering and backend tools, consistently delivering high-performance, scalable solutions.

Top Skills

PythonPyTorchFastapiMongoDBdistilbert

Education

National Institute of Technology, Sikkim

Computer Science and Engineering · Computer Science and Engineering

August 1, 2022 – June 30, 2026

Experience

Premade Innovations Pvt Ltd

AI & ML Intern

June 1, 2025 – August 1, 2025

India

FOXAISR

Full Stack Web Development Intern

January 1, 2025 – March 1, 2025

India

Projects

Controller Load Balancing & Failure Recovery in SDN

January 1, 2025 – June 1, 2026

Designed and implemented the LB-Flex (Load-Balanced Flexible Programmability) framework to optimize controller failure recovery in Software-Defined Networks. Engineered a score-based selection algorithm incorporating a quadratic congestion penalty to prevent controller hotspots and ensure stable load distribution. Evaluated the framework on ATT and BELNET topologies, achieving a 28% reduction in average control-plane latency and a 45% improvement in load distribution fairness compared to baseline greedy algorithms.

View Project

Digital Image Forensic Vision System

January 1, 2024 – June 1, 2026

Engineered a multi-modal forensic classifier integrating 2D Discrete Fourier Transform (DFT) magnitude maps with RGB data to detect high-frequency suppression artifacts unique to modern diffusion models. Fine-tuned an EfficientNet-B0 backbone via Transfer Learning on the 35,000-image GenImage dataset (Midjourney, SD, GLIDE), achieving 95.86% accuracy using AdamW optimization and BCEWithLogitsLoss. Integrated Grad-CAM heatmaps to provide forensic interpretability, visualizing "unnatural" pixel distributions, and deployed the real-time inference dashboard via Streamlit on Hugging Face Spaces.

Multi-Label Toxic Comment Classification

January 1, 2024 – June 1, 2026

Engineered a real-time Transformer-based NLP system utilizing DistilBERT to simultaneously categorize text into 6 toxicity dimensions with <100ms inference latency. Resolved severe class imbalance by implementing stratified sampling and BCEWithLogitsLoss, significantly boosting the model's sensitivity to minority threat categories, deployed via Streamlit.

Certifications

NPTEL Business Intelligence & Analytics (IIT Madras) – Elite, 90% Score

NPTEL (IIT Madras)

June 1, 2026 – Present

NPTEL Computer Vision (IIT Kharagpur) – Elite + Top 1% Scorer

NPTEL (IIT Kharagpur)

January 1, 2025 – Present

GATE 2025 Qualified – AIR 7.9k, Score: 486

Unknown

January 1, 2025 – Present

Key Strengths

Strong foundational knowledge in AI/ML, demonstrated through academic projects and internships.
Proficiency in deep learning frameworks (PyTorch, Transformers) and model architectures (EfficientNet, DistilBERT, Llama 3).
Experience with MLOps and deployment tools (Streamlit, FastAPI, Docker) for real-time inference.
Solid understanding of data preprocessing techniques (stratified sampling, handling class imbalance) and optimization (AdamW, BCEWithLogitsLoss).
Practical experience in RAG systems, multilingual NLP, and computer vision applications.

Cultural & Operational Fit

Cultural Fit Analysis

The candidate's diverse academic projects (computer vision, NLP, SDN) and internships (AI/ML, full-stack web development) indicate a broad interest in technology and a willingness to explore different domains. Their involvement in a hackathon and leadership role suggests a proactive and collaborative mindset. The focus on real-world applications and measurable impact in their projects aligns well with a results-oriented culture. The candidate's continuous learning through NPTEL certifications also points to a growth mindset.

Soft Skills & Operational Fit

The candidate demonstrates strong problem-solving and analytical skills through their project work, particularly in identifying and addressing challenges like class imbalance and controller hotspots. Their experience leading a team in the SIH Hackathon indicates leadership potential and ability to collaborate. The detailed descriptions of project outcomes and methodologies suggest a structured approach to problem-solving and a focus on measurable results. The candidate's ability to deploy applications (Streamlit, FastAPI) shows an understanding of operationalizing ML models.