Senior ML Engineer with 11+ years of experience architecting and deploying scalable AI/ML systems at scale. Expert in LLM-based applications, multi-agent workflows, and distributed systems. Proven track record in technical leadership, system design, and delivering high-impact solutions across conversational AI, NL2SQL, RAG architectures, and agent evaluation frameworks. Adept at mentoring engineers, driving innovation, and building production-grade platforms.

LLMs, Generative AI, Transformers, RAG, NL2SQL, Multi-Agent Workflows, Fine-Tuning (LoRA), Prompt Engineering, BERT, Conversation Analysis

Python, Node.js, Golang, JavaScript, Java, FastAPI, LangChain, PyTorch

Kubeflow, Ray Serve, Databricks, Docker, Kubernetes, MLOps, CI/CD (GitHub Actions), Git

AWS (Lambda, ECS, EC2, S3, RDS), Terraform, Spark, PostgreSQL, MySQL, Elasticsearch, Prometheus, Datadog, Jaeger, Splunk

• M.Tech in Data Science and Engineering - BITS Pilani, 2023
  • Dissertation: Time series forecasting using LSTM with anomaly detection via Isolation Forest for distributed trace analysis
• B.E. in Computer Science Engineering - Anna University, 2014
• Machine Learning Specialization - Coursera (Supervised Learning, Advanced Algorithms, Unsupervised Learning, Recommenders, Reinforcement Learning, Generative AI)

• Architected Activation Agent and Explain My Bill (EMB) Agent within Xfinity Assistant—a multi-agent system where a steering agent orchestrates requests to specialized sub-agents using Dialogflow, AMA, and ADK frameworks
• Designed multi-agent workflow for field technicians combining VAE-based intent classification, multi-modal RAG, Text2SQL, and action decomposition

• Pioneered agent-agnostic evaluation framework supporting ROUGE-based and LLM-as-Judge methodologies across multiple platforms
• Developed grounding verification system to validate chatbot claims against tool responses—published as company-wide Python library
• Built LLM-based conversation simulator replicating diverse user behaviors for comprehensive agent benchmarking

• Architected Kubeflow environment on AWS EKS for ML model deployment with tenant isolation
• Established Ray cluster for distributed training; deployed time series model (FbProphet) via Ray Serve handling 20 concurrent requests/sec
• Built comprehensive LLM observability tool enabling tracing of OpenAI and LangChain calls

• Designed distributed system using consistent hashing algorithm to handle millions of thermostat state change events per minute
• Architected event-driven system using AWS IoT with Lambda functions and API Gateway routing
• Built centralized observability pipeline for 10 microservices incorporating Jaeger, ELK, Prometheus, and Datadog

• 🏆 Recognized with Pinnacle and Spotlight awards for outstanding performance at Comcast
• 🌟 Led Comcast Open Source Hackathon for women engineers
• 🥇 Multiple-time winner at Comcast LabWeek and external hackathons
• 📈 Achieved 92% reduction in processing time (1 hour → 5 minutes for 400 conversations)
• 🚀 Improved activation agent success rate from 55% to 70%

• 📧 Email: abirami.vaishnavi@gmail.com
• 💼 LinkedIn: linkedin.com/in/vaishnaviabiramikumar
• 🐙 GitHub: github.com/vaish1707