Bhavya Geethika

About Bhavya Geethika

Visionary research Leader with 13+ years experience driving analytics and ML innovation across AdTech, marketing science, insurance, and logistics. Deep expertise in experiment design methodologies, attribution modeling, and production-grade systems that power marketing performance and operational optimization at scale. Passionate about leveraging emerging AI to unlock actionable insights, democratize advanced analytics, and empower cross-functional teams to deliver measurable business impact.Hands-on experience building science frameworks that bridge research and production across diverse industries. Proven expertise collaborating with product, marketing, and operations teams on feature development, experimentation infrastructure, and post-launch optimization. Track record of deploying ML/AI innovations for campaign optimization, risk modeling, supply chain analytics, incrementality measurement, and customer lifetime value modeling—consistently driving net incremental revenue growth while improving acquisition, retention, and operational efficiency.**Outside of work, currently focused on building “Breakthroughs Wiki” a Wikipedia like platform that centralizes innovations happening across various fields with a goal of measuring the rate at which progress or stagnation is occurring in fields like cardiovascular & cancer research, robotics, autonomous driving, AI biotech, sports medicine etc.(breakthroughs.today). Technical StackModern GenAI & Agentic Systems • Cursor, Augment Code, Claude Code • LLMops: vLLMs, LangGraph, LangChain, LangSmith, LangFlow, LlamaIndex, Chroma, Qdrant • RAG pipelines, LoRA, QLoRA, RLHF techniques (PPO, DPO), Speculative DecodingLanguages & Frameworks • Python, SQL, R • PyTorch, TensorFlow, Rust •. Tableau, Power BI, Looker, Kibana, DatapineInfra & Databases • Redis, MySQL, MongoDB, Hive, ElasticSearch, Google BigQuery, Vertex AI, AWS Sagemaker, Bedrock, DockerAdvanced ML & AI • GNNs (e.g, GraphSAGE), Transformers (Seq2Seq, Autoregressive) • Reinforcement Learning, LLMs, Generative AI, NLP/NLU/NLG Experimentation & Causal Inference • A/B Testing, Uplift Modeling, Observational Causal Inference methods, Experiment Design

Experience

1. Principal Applied Ai Engineer

   Attadale Partners

   May 2024 — Present · US

   1) Architected AI-Powered Claims Co-Pilot: Architected multi-agent platform processing 60K monthly multimodal insurance claims (video, audio, images, motion modalities), reducing claims specialist review time 82%(45min→8min) at 93% factual accuracy. Designed cloud-agnostic ecosystem using MCP/A2A standards with 15 specialized agents (Whisper ASR, BLIP-2 Vision models (VLM), ImageBind, fraud detection 94% accuracy, LLaMA-2 summarizer fine tuned on Q-LoRA). Deployed 5 adjacent workflows in 9-14 days vs. 4-6 months (83% reduction) with 65% agent reuse across 308K monthly transactions. Implemented Not Diamond (multimodel infrastructure for routing to premium vs cheaper models) routing (58% cost reduction), automated testing (73% failed tests reduction), and validated Kubernetes portability.2) Implemented AI-Driven Risk Model: Deployed ML risk assessment processing 1M+ annual applications, reducing loss ratio 10 points (75%→65%) for $8M annual savings. Built ensemble with 200+ features achieving 0.85 AUC, 82% recall, enabling 70% auto-decisioning <10 seconds at 99.9% uptime.3) Built Product Concept Testing Platform: Architected multi-tenant Django application with Azure OpenAI/DALL·E for rapid prototyping, maintaining strict client data isolation across 50+ concepts. Developed chatbot with 200 digital twins achieving 12% MAPE, RAG retrieving 25K+ historical launches, contributing to $2M+ agency business at 85% adoption.4) Built Census Cognitive Map Platform: Architected LLM analytics leveraging 25K+ tables with LangChain orchestration, reducing analysis time from 3 hours to 22 minutes, handling 200+ concurrent users across 4,500 conversations with 97.2% context accuracy.5) Engineered Azure AI Search Platform: Deployed vector-based semantic search processing 10K+ documents, reducing retrieval from 15min to 2min, improving accuracy 60%→85%, achieving 450ms p95 latency at 500+ concurrent users with 99.94% uptime.

Education

• Vellore Institute of Technology

  Bachelor's Degree, Electrical and Electronics Engineering

• University of Illinois Chicago

  Master's degree, Management Information Systems and Services

Skills

• Marketing
• Microsoft Office
• Matlab
• Python
• Selenium
• Strategic Planning
• Microsoft Excel
• Consulting
• Business Intelligence
• Management
• Html
• Databases
• Data Mining
• Requirements Analysis
• Editorial Calendar
• Web Analytics
• Machine Learning
• Social Network Analysis (Sna)
• C++
• SEO
• Project Management
• C
• Java
• Sql
• Social Network Analysis
• Adobe Muse
• Analytics
• Google Adwords
• Analysis
• Search Engine Optimization (SEO)
• Javascript
• Bundled Payment
• R
• Data Analysis
• Squarespace
• Strategy
• Google Analytics
• Leadership
• Testing