Manogna Venkata Sudha Anirudh Maganti
About
Detail
AI/ML Engineer | Data Scientist | Machine Learning, Deep Learning, NLP, LLMs, SQL, Python
Indiana, United States
Résumé
Jobs
verified_user
0% verified
AI/ML EngineerADPApr 2024 - Current (2 years 2 months)Engineered an LLM-powered HR assistant using GPT-4 + LangChain with RAG (Pinecone/FAISS), implementing NLP models with entity recognition, intent classification, and sentiment analysis to improve query resolution by 30% and reduce escalations by 25%. Deployed advanced churn prediction pipelines on SageMaker using predictive modeling techniques, achieving 20% improvement in accuracy for employee retention forecasting and automating retraining with Feature Store + CI/CD workflows. Engineered real-time ETL pipelines using Apache Spark and Kafka to process transactional HR and payroll data for enterprise clients, enabling immediate access to employee performance metrics and reducing data retrieval time by 40%. Implemented time series forecastin- AI/ML EngineerTaylor CorporationFeb 2023 - Jan 2024 (1 year)Developed GenAI-powered recommender systems using LangChain + RAG (Pinecone) with content-based filtering and NLP techniques (sentiment analysis, topic modeling via NLTK, SpaCy, Gensim), boosting user engagement by 20% through personalized content delivery. Architected end-to-end multimodal ML pipelines using PyTorch + PySpark + MLflow with Flask RESTful APIs, streamlining model deployment by 25% and enabling rapid iteration for product teams and stakeholders. Built real-time conversational AI systems using Azure Bot Service + LangChain, integrating NLU capabilities with Azure Cognitive Services and RAG models to improve response accuracy by 30% and reduce query resolution time through contextually relevant answers. Managed large-scale data
- HData ScientistHyundai Motor India Engineering Pty LtdJul 2019 - Dec 2022 (3 years 6 months)Implemented time-series demand forecasting with Python + SQL, improving production planning accuracy by 13%. Applied simulation-driven design analytics to evaluate component reliability using past model data and virtual testing, accelerating product development and reducing prototype testing cycles. Delivered ML-driven defect detection systems on assembly lines using sensor data, minimizing quality control failures and optimizing inspection workflows. Designed predictive maintenance models for manufacturing machinery using IoT data and historical downtime trends, boosting equipment uptime and reducing unplanned stoppages. Led integration of sensor fusion pipelines (cameras, vehicle telemetry, ADAS data) to drive next-gen safety and autonomy
Education
verified_user
0% verified
- Master of SciencePurdue UniversityJan 2023 - Dec 2024 (2 years)Major: Computational Data Science (Computer Science & Mathematics)GPA: 3.5/4.0, Dean's Scholarship Recipient
- Bachelor of TechnologyGayatri Vidya Parishad College of EngineeringJun 2015 - May 2019 (4 years)Major: Mechanical EngineeringCGPA: 8.63/10.00, Full-ride scholarship
Projects (professional or personal)
verified_user
0% verified
- BBook Recommender SystemDesigned a hybrid engine on 10M+ interactions using SVD matrix factorization and TF-IDF, achieving 87% precision@10 for personalized book recommendations.
- RRevenue Forecasting (Time Series)Built ARIMA and Prophet models on 5 years of e-commerce data, achieving 8.5% MAPE for monthly revenue prediction and supporting quarterly financial planning.
- LLLM from ScratchImplemented a transformer-based language model from scratch in PyTorch with multi-head attention mechanisms, positional encodings, and layer normalization, achieving perplexity score of 45.2 on text completion tasks through custom tokenization and gradient optimization strategies.
- AAgentic AI for Data AnalysisDeveloped an autonomous AI agent using LangChain Agents + OpenAI GPT-4 to orchestrate complex analytical workflows (SQL querying, REST API integration, and automated visualization), achieving 95% accuracy in automated insights generation.
- BBayesian Health Insurance Cost PredictionConstructed a Bayesian model using MCMC for predicting health insurance costs; achieved R-squared of 0.89 and RMSE of 0.46 using Python and R.
- CClick-Through Rate Prediction | Marketing Campaign OptimizationFormulated a click-through rate prediction model on highly imbalanced marketing data (~1.6% positive rate), optimizing for log loss and calibrated probabilities rather than raw classification accuracy, ensuring reliable ranking of high-intent users.
- CCrop row and Leaf Segmentation (Deep Learning)Constructed U-Net models achieving 85% accuracy in crop row detection and 97% in leaf segmentation, enabling precise yield assessment and optimized planting.
- MMetaphor Detection (NLP)Trained a DistilBERT-based model to detect metaphors in text, achieving a state-of-the-art NLU loss of 0.045 using Python and TensorFlow.