Vaibhav Tej

About Vaibhav Tej

Senior AI/ML professional with 8+ years across Machine Learning and Data Science, and architecting enterprise-grade Generative AI, LLM, and RAG systems across finance, banking and healthcare domains.Built multimodal AI pipelines using CNNs, Vision Transformers, U-Net, CLIP-style encoders, OCR, and ASR models for document intelligence, radiology workflows, and content extraction.Strong ML engineering foundation with PyTorch, TensorFlow, scikit-learn, XGBoost, LightGBM, delivering accurate classification, forecasting, anomaly detection, and NLP models. Expert in building LLM applications using LangChain, LangGraph, LlamaIndex, Hugging Face, GPT-4/4o, Gemini 1.5, and Claude 3, improving automation accuracy and reasoning quality.Architected advanced RAG systems using Pinecone, FAISS, Weaviate, Chroma, hybrid retrieval, and custom embeddings, enabling high-precision enterprise search and knowledge intelligence.Skilled in LLM fine-tuning with LoRA, QLoRA, PEFT, optimizing foundation models for financial, clinical, and compliance-heavy use cases while reducing compute cost.Developed multi-agent orchestration workflows using LangGraph and CrewAI for tool-calling, validation agents, automated reasoning, and workflow supervision.Expert in Machine Learning, Deep Learning, and Generative AI (LLMs) using Python, TensorFlow, PyTorch, Hugging Face, and OpenAI GPT APIs.Integrated scalable LLM microservices with REST APIs, FastAPI, gRPC, Serverless, API Gateway, achieving sub-second inference and high-availability deployments.Implemented observability and AI safety using CloudWatch, Prometheus, OpenTelemetry, guardrails, hallucination detection, toxicity filters, and response-quality scoring.Proficient in Python, TensorFlow, PyTorch, and Scikit-learn, with deep hands-on experience fine-tuning LLMs (GPT, LLaMA, Falcon) for text generation, reasoning, and task planning.

Experience

1. Gen Ai Engineer

   高盛

   Sep 2024 — Present · Dallas, TX, US

   Project: AI/ML & Generative AI Modernization Program, Developed and deployed enterprise-grade GenAI, LLM, and RAG systems to modernize search, automation, and decision workflows across financial operations. Built scalable LLM applications, multi-agent pipelines, and cloud-native architectures, ensuring high accuracy, low latency, and compliance in regulated environments.Responsibilities:· Architected end-to-end RAG pipelines using LangChain, LangGraph, Pinecone, and OpenAI embeddings, improving knowledge-retrieval precision by 35% and reducing regulated-query hallucinations by 40%.· Built hybrid retrieval engines combining vector search + BM25 with metadata filters and re-ranking, increasing financial document recall for audit teams by 28%. · Built automated ingestion pipelines with AWS Glue, Lambda, S3, processing thousands of financial documents daily with lineage tracking and deterministic versioning. · Implemented multimodal LLM workflows using AWS Transcribe + GPT-4o to auto-generate call summaries, reducing analyst review time by 60%. · Deployed secure RAG inference stacks using Lambda, API Gateway, DynamoDB, delivering under 700ms latency for internal enterprise search operations.

Education

• Computer science

  Computer science