Vrunda Teeleru

Hi, I'm Vrunda,

Most AI engineers build things that work in demos. I specialize in making them work when the stakes are real: compliance audits, high-net-worth clients, fraud catching at 2,000 transactions per second. What I bring that is rare is the ability to go backward, from a broken output to the exact field, join, or retrieval failure that caused it. I have done that across banks, broker-dealers, and telecom data at serious scale. And outside of work, I built an entire AI product completely solo, because I wanted to prove to myself that I could.

"

ARTIFICIAL INTELLIGENCE IS NOT A SUBSTITUTE FOR HUMAN INTELLIGENCE; IT IS A TOOL TO AMPLIFY HUMAN CREATIVITY AND INGENUITY

Work Experience

Where I've shipped real production AI

01
LPL Financial
Generative AI Engineer
Own the MLOps lifecycle for LLM and RAG pipelines. Redesigned a compliance RAG, lifting precision@5 from 64% to 89%. Built LangChain agents auto-resolving ~55% of routine queries.
02
Fisher Investments
Machine Learning Engineer
Owned SageMaker deployment lifecycle for client analytics. Built dual-signal churn model flagging at-risk HNW clients ~5 weeks earlier (AUC 0.78 → 0.86). Diagnosed a feature-join bug affecting 12% of clients.
03
Accenture — AT&T
Data Scientist
NLP incident classification on telecom logs at 87% F1, ~40K tickets/month. Real-time anomaly detection on Kafka/Spark over CDRs cut detection lag from ~6 hours to under 90 seconds.
04
Kotak Mahindra Bank
Data Scientist
Diagnosed silent model degradation across 3 core banking systems and ~20 conflicting fields. Built a BERT + XGBoost KYC document classifier at 94% accuracy, clearing ~3,000 docs/day.
Skillset

Numbers that hold up under audit

Five years of production AI and ML systems in regulated finance and at telecom scale. The work isn’t the benchmark; it’s the metric that survives a model risk review.

64→89%
Compliance RAG retrieval precision@5
Generative AI · RAG
3.7×
LLM inference speedup via ONNX Runtime port
MLOps · Inference
~55%
Routine compliance queries auto-resolved by LangChain agents
Agentic AI · LangChain
94%
KYC document classification accuracy (BERT + XGBoost)
Deep Learning · NLP
87% F1
NLP incident classification across ~40K monthly tickets
NLP · Production
~2,000/sec
Fraud detection throughput on Kafka & Spark streaming
Real-Time ML · Streaming
Python
PyTorch
TensorFlow
Hugging Face
LangChain
LlamaIndex
OpenAI
Anthropic
FAISS
Pinecone
ChromaDB
ONNX
BERT
GPT
LayoutLMv3
SpaCy
Scikit-learn
XGBoost
LightGBM
Apache Kafka
Apache Spark
Airflow
Databricks
SQL
aws AWS
SageMaker
Google Cloud
Vertex AI
Docker
Kubernetes
MLflow
Kubeflow
FastAPI
GitHub Actions
Pandas
NumPy
Jupyter
Tableau
Power BI
Python
PyTorch
TensorFlow
Hugging Face
LangChain
LlamaIndex
OpenAI
Anthropic
FAISS
Pinecone
ChromaDB
ONNX
BERT
GPT
LayoutLMv3
SpaCy
Scikit-learn
XGBoost
LightGBM
Apache Kafka
Apache Spark
Airflow
Databricks
SQL
aws AWS
SageMaker
Google Cloud
Vertex AI
Docker
Kubernetes
MLflow
Kubeflow
FastAPI
GitHub Actions
Pandas
NumPy
Jupyter
Tableau
Power BI
My Projects

AI driven Projects

Snapseek

Live AI photo platform for events. Guests take a selfie and get every photo they appear in, under 3 seconds. Built solo across ML, backend, frontend, and infrastructure — serving dozens of real events including full wedding scale. Parallel face detector ensemble, p50 of 1.1s.

Visit Project →

Financial RAG Evaluation

Production-style RAG over 5 years of SEC filings (6,447 chunks). 4 retrieval strategies, FastAPI A/B routing, McNemar's test on a 514-question golden set. Headline finding: domain representation in the corpus, not retrieval strategy, was the real accuracy bottleneck.

View on GitHub →
Education

Schools I have learnt from in the past

01

Texas A&M University,
M.S. in Computer Science

Completed a Master's in Computer Science in the U.S., gaining advanced grounding in algorithms, machine learning, and data systems that translated directly into production work in regulated finance.

02

Sri Padmavathi Mahila VV,
B.S. in Computer Science

Completed a B.S. in Computer Science, focusing on algorithms, data structures, and software engineering. Built the technical foundation for a career in AI and ML systems.

03

Certifications,
AWS & Azure

AWS Certified Machine Learning — Specialty. Microsoft Certified: Azure AI Engineer Associate. Continuous learning across the cloud and ML specialty stack that backs my daily production work.

Professional Networking

My LinkedIn connections are established with professionalism and a commitment to mutual growth.
Connect with me for more brainstorming.

Connect on LinkedIn

Let's build something real.

Open to senior IC roles in Generative AI, applied ML, and ML infrastructure. Especially interested in work involving real production systems and real stakeholders.

vrundareddyteeleru09@gmail.com LinkedIn GitHub +1 (361) 742-5890