Karthik A
About
Detail
AI & ML Engineer at Deloitte USI
Bengaluru, Karnataka, India
Contact Karthik regarding:
work
Full-time jobs
Résumé
Jobs
verified_user
0% verified
Senior AI EngineerKnowBe4Nov 2025 - Current (11 months)Working on Transformer-based NLP models with a strong focus on model optimization and high-performance inference using ONNX, Optimum, and TensorRT. Experienced in building end-to-end optimization pipelines, including:
• Exporting Transformer models to ONNX
• Applying advanced graph optimizations (operator fusion, constant folding, transformer-specific optimizations)
• Precision optimization using FP16 and INT8 (CPU/GPU-aware strategies)
• Backend-specific tuning for ONNX Runtime and TensorRT Deploying and scaling optimized models on AWS SageMaker and NVIDIA Triton Inference Server, targeting NVIDIA GPUs, including:
• NVIDIA A10G (Ampere architecture) — used in AWS G5 instances
• NVIDIA V100 (Volta architecture)- Senior AI and ML EngineerDeloitteApr 2025 - Nov 2025 (8 months)AI Engineer | Multi-Agent LLM Architect | Claims AI | AWS SageMaker Expert | AI Observability I design and deploy intelligent multi-agent systems using cutting-edge frameworks like LangGraph, LangChain, CrewAI, GPT-4o, and Meta LLaMA to analyze and resolve complex healthcare claims inquiry data. Key areas of focus:
• Fine-tuned Meta Llama-3.2-1B using QLoRA on AWS SageMaker (g5 GPU) using a 150k domain dataset for instruction-following, text generation and domain-specific Q&A. Optimized the model using PEFT, BF16, and 4-bit quantization to reduce GPU cost while achieving strong accuracy and lower inference latency.
• Building multi-agent architectures powered by large language models
• Integrating vector databases for efficie - Senior AI and ML EngineerCARELONApr 2022 - Mar 2025 (3 years)Project 1: Smart Analyzer for Inquiries & Claims Processing. Overview: Developed an NLP-based system to streamline inquiry and claims processing, improving efficiency and user satisfaction. Objectives: - Enhance efficiency in handling inquiries and claims. - Achieve significant cost savings and revenue generation. - Improve user satisfaction for members, providers, and associates. Description: Utilized NLP and NER to analyze inquiry notes and claim data, determining intent and recommending actions to ensure accurate claim settlements and reduce future adjustments. Technical Skills: - NLP (SpaCy, NLTK, BERT, DistilBERT) - Data Cleaning - Keyword Extraction - Clustering (K-Means, Soft-clustering, HDBSCAN, Agglomerative,) - Text generation & S
- AI Engineer and ML EngineerUSTMar 2021 - Apr 2022 (1 year 2 months)Project Title: AI-Driven Healthcare Claims Adjudication System Overview: This system revolutionizes claims processing using AI/ML, NLP, and Generative AI, enhancing efficiency, accuracy, and fairness while reducing manual intervention. Objectives: 1. Efficiency: Streamline workflow to reduce processing time. 2. Accuracy: Improve claims processing accuracy with advanced AI models. 3. Fairness: Ensure fair decisions by addressing biases and implementing human-in-the-loop workflows. Responsibilities: - Vector Database Design: Optimize databases for efficient data retrieval. - RAG Pipelines: Develop and optimize Retrieval-Augmented Generation pipelines. - Langchain Integration: Enhance adaptability and precision of Generative AI models. - Human
- Machine Learning EngineerCognizantSep 2018 - Sep 2020 (2 years 1 month)Project Title: Walmart Retail Accounts Payable Analysis and Optimization Overview: Analyze and optimize Walmart's accounts payable process using data science techniques to predict payment delays, identify fraudulent transactions, and optimize payment schedules. Objectives: 1. Predict payment delays with machine learning. 2. Detect fraudulent transactions using anomaly detection. 3. Optimize payment schedules to improve cash flow. Data Sources: - Accounts payable: Historical invoices, payment dates, amounts, vendors. - Vendor data: Credit terms, reliability, transaction history. - Financial data: Walmart's cash flow patterns. Techniques and Skills: 1. Data cleaning and preprocessing. 2. Exploratory data analysis (EDA) with Matplotlib and Sea
Education
verified_user
0% verified
- bachelor, Bachelor of Engineering - BE Electronics and Communications EngineeringVisvesvaraya Technological UniversityJan 2014 - Jan 2018 (4 years 1 month)