Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

Machine Learning Engineer with 4+ years building and deploying production AI/ML systems, specializing in Generative AI, LLMs, and RAG architectures for financial services. Proven impact through fine-tuning LLaMA 3 (LoRA/QLoRA), building multi-agent pipelines with LangGraph, and optimizing LLM inference with vLLM and quantization to reduce costs and meet high uptime and latency SLAs.…

Niharika Amilineni

AI Engineer, Data Scientist, Programmer, +4





Machine Learning Engineer with 4+ years building and deploying production AI/ML systems, specializing in Generative AI, LLMs, and RAG architectures for financial services. Proven impact through fine-tuning LLaMA 3 (LoRA/QLoRA), building multi-agent pipelines with LangGraph, and optimizing LLM inference with vLLM and quantization to reduce costs and meet high uptime and latency SLAs.…

Available to hire

Machine Learning Engineer with 4+ years building and deploying production AI/ML systems, specializing in Generative AI, LLMs, and RAG architectures for financial services.

Proven impact through fine-tuning LLaMA 3 (LoRA/QLoRA), building multi-agent pipelines with LangGraph, and optimizing LLM inference with vLLM and quantization to reduce costs and meet high uptime and latency SLAs.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Intermediate

Language

English

Advanced

Work Experience

Machine Learning Engineer at S&P Global

October 1, 2024 - Present

Architected an enterprise RAG-based financial intelligence platform using Python, LlamaIndex, and Pinecone, implementing semantic chunking and hybrid search to reduce analyst research turnaround time by 28%. Developed multi-agent AI workflows using LangGraph with tool-calling and self-correction loops to automate complex document analysis and eliminate manual risk compliance workflows. Fine-tuned LLaMA 3 with LoRA/QLoRA on proprietary financial corpora using Hugging Face and Weights & Biases, integrating DeepSpeed for multi-GPU training to improve entity extraction accuracy by 22%. Optimized inference via AWQ/INT8 quantization and structured pruning, deploying with vLLM to cut inference costs by 40% and achieve sub-150ms P99 latency. Engineered scalable LLM infrastructure on AWS EKS and Kubeflow, enabling 10,000+ daily evaluations with 99.5% uptime. Built production LLMOps pipelines integrating TruLens for evaluation and Arize for drift monitoring, automating CI/CD with Docker and Terr

Machine Learning Engineer at Freshworks

June 1, 2020 - July 1, 2023

Productionized fraud and risk detection services using Random Forest and LightGBM, reducing false alert rates by 22% and delivering over $1M in annual cost savings through optimized inference and threshold tuning. Deployed an AWS-native ML pipeline using S3, SQS, EC2, and SageMaker for real-time scoring across 500K+ daily records, cutting infrastructure overhead by 28% while sustaining 99.8% SLA. Designed unsupervised anomaly detection systems using Isolation Forest and Autoencoders to uncover 2.5M in hidden fraudulent activity. Built statistical A/B testing frameworks with Python and StatsModels to evaluate 35+ models, improving detection recall by 15%. Optimized time-series forecasting pipelines using SQL, Python, ARIMA, and Prophet, improving prediction accuracy by 30% via automated feature engineering and retraining loops. Established an MLOps framework using MLflow and Kubeflow for tracking, reproducibility, and governance, accelerating compliance approval for new models by 25%.