Hello, I'm Vishnu sai

I'm a

Welcome to my personal website where I showcase my work and skills.

Contact Me View My Work
Portrait of vishnu sai

About Me

Get to know me!

Vishnu Sai is a Machine Learning Engineer, Deep Learning Enthusiast,and developer—constantly evolving and passionate about learning new frontiers.

I specialize in creating intelligent systems and data-driven solutions. With a background in Machine Learning, I combine analytical thinking and technical skills to drive business.

When I'm not working, you can find me exploring new technologies, contributing to open-source projects, playing pickle ball or reading books.

Resume

Education

Master's in Applied Machine Learning, UMD (2025-2027)

Bachelor's in Computer Science(Data Science), VIT (2019-2023)

Location

College Park, Maryland - United States

Languages

English, Hindi, Telugu

Experience

Machine Learning Engineer

Jun 2023 – Jul 2025

Maruti Suzuki India Limited · Bengaluru, India

  • Agentic Data Analyst: Automated data analysis by architecting and building a multi-agentic system using Agency Swarm and Azure OpenAI API.
  • Engineered memory using Memgraph for short-term context graphs and ChromaDB + Neo4j for long-term RAG retrieval.
  • Enhanced contextual reasoning with LangChain pipelines, achieving 99% retrieval accuracy and 30% lower token usage.
  • Built a Streamlit interface that reduced analysis time from 10 days to 10 minutes and cut manual effort by 90%.

  • Telematics Chatbot – Natural Language to SQL: Designed a GenAI chatbot to democratize data access by translating natural language into optimized SQL for self-service analytics.
  • Fine-tuned Flan-T5 using LoRA and PEFT on domain SQL/NL datasets, achieving 99% accurate query-to-database mapping.
  • Optimized inference using ONNX Runtime and mixed-precision quantization, reducing GPU memory by 25% and latency by 40%.
  • Drove business impact by boosting user engagement by 80% and improving cross-team collaboration by 58%.

Machine Learning Intern

Jan 2023 – Jun 2023

Maruti Suzuki India Limited · Bengaluru, India

  • Accident Detection using LSTM Autoencoders: Developed an LSTM autoencoder to predict accidents by modeling sequential time-series data and detecting anomalies.
  • Engineered features using rolling statistics and window-based methods to capture temporal patterns and boost accuracy.
  • Validated insights through EDA and hypothesis testing, achieving an F1-score of 82% for deployment readiness.
  • Created a benchmark dataset from processed accident data to enable hotspot dashboards and improve safety design.

My Projects

My Skills

Programming & Data

Python

Advanced

SQL

Advanced

Modeling & Machine Learning

TensorFlow

scikit-learn

pytorch

CUDA

Transformer

AWS

Azure

OpenAI

MLOps & Deployment

MLflow

MLX

Docker

FastAPI

Professional Skills

Communication

Teamwork

Problem Solving

Creativity

Tools & Platforms

Git Git
Docker Docker
Jupyter Jupyter
VS Code VS Code
Cursor Cursor

Contact Me

Get In Touch

Feel free to reach out to me for any questions or opportunities!

Email

vishnubr@umd.edu

vishnupuma18@gmail.com

Location

College Park, Maryland

(open to relocation)
Python ML Code Sample 

 from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
import torch

# -----------------------------
# MODEL SETUP
# -----------------------------
print("🚀 Vishnu Projects: Initializing EmotionSense Model...\n")

MODEL_NAME = "j-hartmann/emotion-english-distilroberta-base"
emotion_pipeline = pipeline("text-classification", model=MODEL_NAME, return_all_scores=True)

# -----------------------------
# FUNCTION FOR INFERENCE
# -----------------------------
def analyze_emotion(text: str):
    """
    Analyze the emotion of the given text using the Transformer model.
    """
    results = emotion_pipeline(text)[0]
    # Sort by highest score
    results = sorted(results, key=lambda x: x["score"], reverse=True)
    top_emotion = results[0]
    print(f"🧩 Text: {text}")
    print(f"💬 Predicted Emotion: {top_emotion['label']} ({top_emotion['score']:.3f})\n")

# -----------------------------
# DEMO INPUTS
# -----------------------------
if __name__ == "__main__":
    print("🔍 Running Emotion Analysis with Transformers...\n")

    samples = [
        "I just got promoted and I feel amazing!",
        "I'm so tired of everything going wrong lately.",
        "I miss my friends back home.",
        "Wow, this AI project is mind-blowing!"
    ]

    for s in samples:
        analyze_emotion(s)

    print("✅ Vishnu Projects – EmotionSense completed successfully.")