Resume
Available for GenAI & AI/ML Opportunities

Engineering Intelligent Systems That Make An Impact

Specializing as

Hi, I'm Sivanesan B. I bridge cutting-edge AI research with production-grade software engineering. From authoring IEEE-published carbon-efficient LLM inference techniques to building autonomous agent workflows, voice AI platforms, and real-time computer vision systems.

Explore My Creations IEEE Publication
0%
LLM CO2 Reduction
0%
Faster Inference
0%
Vision Model Acc.
0%
Screening Time Cut
Sivanesan B - AI Engineer
PyTorch PyTorch
LangChain LangChain
FastAPI FastAPI
Hugging Face Transformers
// BIOGRAPHY & VISION

Driven by Curiosity, Engineered for Scale

Transforming complex AI research and models into reliable, high-utility products.

Who I Am

I am an AI Engineer currently working as a GenAI Executive & Prompt Engineer at Nunes Instruments. My passion lies at the intersection of Generative AI, Large Language Models, autonomous agent architectures, and real-time computer vision.

I believe world-class AI goes beyond just having high test set accuracy. It must be computationally efficient, privacy-conscious, easy to operate in low-resource environments, and engineered with seamless user experiences.

Coimbatore, Tamil Nadu, India
B.Tech in AI & Data Science (KCT)
Honor & Recognition

Best Final Year Project Award

Awarded by KCT Department of AI & DS (Aug 2025) for research on Carbon-Efficient LLM Inference.

58%
Lower LLM Carbon Footprint
56%
Inference Speedup via Hook Pruning
92%
ConvNeXt Real-World Accuracy
<50ms
Low-Latency Vision Proctoring (30 FPS)
// RESEARCH & SCHOLARSHIP

Peer-Reviewed Publications

Advancing carbon-efficient, sustainable deep learning inference for modern LLMs.

Eco-Transformers Paper Cover
Sustainable AI IEEE ICNGCS-2025

Eco-Transformers: Carbon-Efficient and Accelerated Inference through Self-Pruning and Activation Freezing in Large Language Models

Sivanesan B, Shriaarthy E, Sangeetha S, Niranjana R S, Rohini R, Vijayalakshmi N

Conference: 3rd International Conference on Next Generation Computing Systems (ICNGCS - 2025)
Venue & Date: PSG iTech, Coimbatore, India | August 21–22, 2025
Publisher & DOI: IEEE | 10.1109/ICNGCS64900.2025.11183217

Large Language Models require immense computational power, leading to high latency and heavy carbon emissions. We introduce Eco-Transformers, a novel runtime optimization framework implementing PyTorch forward hooks that dynamically masks low-magnitude weights and freezes near-zero activations at runtime without requiring retraining or modifying model architecture. Evaluated on Falcon-RW-1B, GPT-Neo-1.3B, OPT-1.3B, and GPT2-XL, Eco-Transformers consistently reduced CO2 emissions by up to 58% and accelerated inference by 56% while preserving BLEU, ROUGE-L, and perplexity scores.

View on IEEE Xplore
// SELECTED WORKS

Featured AI Creations

Production-ready AI agents, vision pipelines, voice intelligence, and fine-tuned models.

WhatsApp Intelligence System Enterprise GenAI

WhatsApp Product & Price Intelligence

Autonomous enterprise GenAI system deployed to parse unstructured customer conversations, extracting product names, prices, and quotation details into structured SQLite archives with automated management reports.

  • Unstructured chat & document information extraction
  • Automated product catalog & pricing reconciliation
  • Fast historical search archive with Flask UI
Python Prompt Engineering SQLite Flask Document AI
Repository
TestPilot AI Platform Autonomous Agents

TestPilot AI – Autonomous Test Engineer

A comprehensive autonomous AI software quality platform that performs intelligent UI testing, automated API testing, accessibility auditing, security analysis, and AI-powered root-cause bug diagnosis.

  • Autonomous multi-agent orchestration for QA
  • Intelligent bug reports & root cause analysis
  • Automated security & accessibility scanners
Python LangChain Groq PostgreSQL AI Agents
GitHub Code
AI Book Scanner Mobile App 92% Detection Acc

AI Book Scanner (ConvNeXt + Vision LLM)

Flutter mobile application that recognizes books from camera feeds using a custom fine-tuned ConvNeXt model (92% accuracy across 1,000+ test images) with FAISS semantic retrieval and Vision LLM contextual summaries.

  • Fine-tuned ConvNeXt transfer learning on PyTorch
  • FastAPI microservice built for 100+ concurrent requests
  • FAISS vector index for lightning-fast book discovery
Flutter ConvNeXt FastAPI PyTorch FAISS Vision LLMs
APK Demo
AI Recruiter Voice Interview Platform Voice AI + Vision

AI Recruiter – Smart Voice Interviewer

Real-time voice interview platform powered by Vapi AI and OpenRouter LLMs. Conducts dynamic interviews, automatically evaluates responses across 50+ mock sessions, and runs BlazeFace proctoring at 30 FPS under 50ms latency.

  • Low-latency conversational voice interview flows
  • BlazeFace proctoring running smoothly in-browser
  • Reduced manual HR screening overhead by 70%
Next.js Vapi AI OpenRouter BlazeFace Supabase
GitHub Code
Offline Voice Assistant 100% Offline AI

Offline Voice Assistant (Whisper + Qwen)

Privacy-first on-device voice assistant for warehouse inventory. Runs local Whisper STT, FAISS vector retrieval, and quantized Qwen language model without needing an active internet connection.

  • Zero internet dependency for air-gapped security
  • Quantized on-device LLM running low latency
  • Slashing manual stock lookup duration by 45%
Python Whisper STT FAISS RAG Qwen 1.5
GitHub Code
LLM Fine-Tuning LoRA / QLoRA

Fine-Tuning Qwen1.5-0.5B for Custom Chat

Domain adaptation and fine-tuning of the Qwen language model using Hugging Face Transformers Trainer API and parameter-efficient techniques for custom bot personas and consistent formatted answers.

  • Custom instruction-tuning dataset curation
  • Hugging Face Trainer API parameter optimization
  • Evaluation benchmarks for tone and consistency
Python Hugging Face Transformers LoRA
GitHub Code
// COMPETENCIES

Core Technical Toolkit

Technologies, frameworks, and architectures I leverage daily.

Generative AI & LLMs
LLMs & VLMs LangChain & LangGraph RAG & Vector DBs AI Agent Workflows FAISS / ChromaDB LoRA / QLoRA Prompt Engineering
Deep Learning & Vision
PyTorch TensorFlow OpenCV ConvNeXt Whisper STT NumPy Pandas
Backend & Cloud
FastAPI Flask Docker Google Cloud Supabase Firebase PostgreSQL SQLite
Languages & Tools
Python TypeScript C++ Git & GitHub Linux Postman Flutter

Professional Certifications

IBM AI Engineering Professional
Coursera / IBM Deep Learning & AI
AWS AI & ML Scholars
Udacity & Amazon Web Services
Tata GenAI Job Simulation
Forage / Enterprise AI Systems
// TIMELINE

My Journey & Roles

Key milestones shaping my engineering perspectives and achievements.

AUG 2026 – PRESENT

GenAI Executive & Prompt Engineer

Nunes Instruments
  • Developing LLM-based business automation applications and GenAI pipelines to process unstructured business data.
  • Designing, benchmarking, and evaluating prompts for structured information extraction, classification, and summarization.
  • Architecting multi-agent workflows connecting LLMs with core business databases and REST APIs for real-time operations.
JUL 2026 – AUG 2026

AI Coder (Graduate Apprentice)

Madhu Jayanti International Pvt. Ltd.
  • Engineered Python data automations using openpyxl to process high-volume operational metrics into structured shift reports.
  • Automated IoT production machine monitoring to record manufacturing cycles and generate live audit reports.
  • Configured automated WhatsApp and email notification alerts to designated team leads, cutting reporting lag to zero.
JAN 2025 – JUL 2025

Innovation Engineer Trainee

Forge Innovation & Ventures, Coimbatore
  • Built an offline voice-driven inventory assistant integrating Whisper STT, on-device quantized Qwen LLM, and FAISS RAG.
  • Reduced manual stock-lookup time by 45% across 3 warehouse pilot deployments with zero internet connectivity.
  • Engaged in rapid hardware-software prototyping and startup incubation methodologies.
2023 – 2026

B.Tech in Artificial Intelligence & Data Science

Kumaraguru College of Technology (KCT), Coimbatore | CGPA: 8.02 / 10
  • Honored with the Best Final Year Project Award for research in carbon-efficient LLM inference.
  • Published IEEE paper on Eco-Transformers runtime activation freezing and pruning.
JUN 2022 – JUL 2022

Data Networking & Infrastructure Intern

Bharat Sanchar Nigam Limited (BSNL)
  • Gained hands-on exposure to high-capacity optical fiber networks, node routing, and telecommunication switching systems.
  • Studied network latency, throughput optimization, and real-time telecom transmission architectures.
2020 – 2023

Diploma in Computer Engineering

Tamilnadu Government Polytechnic College, Madurai | CGPA: 8.0 / 10
  • Core foundational training in C, Python, web servers (Apache), Linux systems, and IoT networking.
// ARTICLES & WRITINGS

Insights & Reflections

Breaking down complex AI concepts with memorable analogies and technical depth.

Neural Networks & Catastrophic Forgetting AI Fundamentals
AUG 5, 2025

Why Neural Networks Forget Like Humans—And How We Help Them Remember

Drawing inspiration from cinema and neuroscience, this deep dive explains why neural networks suffer catastrophic forgetting when learning sequentially, and how dual-network architectures preserve lifelong memory.

Read Full Article
Neural Architecture Search AutoML & NAS
AUG 6, 2025

AI Models That Build Themselves: The Rise of Neural Architecture Search

Exploring how automated machine learning and evolutionary reinforcement strategies enable algorithms to design their own optimal neural topologies.

Read Full Article
Browse All Articles & Insights
// GET IN TOUCH

Have an idea? A challenge?
Let's build something remarkable together.

Whether you want to discuss GenAI opportunities, LLM efficiency research, custom agent development, or collaborative projects, my inbox is always open.

Send Email (apsiva69@gmail.com) +91-8438692752
Siva AI Assistant
● Online & Ready
Hello! I'm Sivanesan's Portfolio AI. Ask me anything about his Eco-Transformers research paper, autonomous AI projects, GenAI work at Nunes Instruments, or how to get in touch!