Parthu Maddhuri

Proof of Experience

Value Labs · Dec 2025 – Present

Engineered real-time SageMaker inference APIs via API Gateway + Lambda achieving P95 latency <250ms, reducing operational overhead by 30% versus prior batch-only architecture.
Deployed DeepAR + XGBoost ensemble forecasting for volunteer demand prediction, cutting manual scheduling effort by 25% and improving resource allocation accuracy by 38%.
Built PySpark ETL on AWS Glue with PII tokenization, deduplication, and schema normalization processing 10M+ daily records in full GDPR/ISO compliance.
Delivered 40% analytics speed improvement by migrating reporting to AWS Athena federated queries over S3 data lake, eliminating ETL bottlenecks for BI workloads.
Implemented production model drift monitoring (PSI/KL-divergence + CloudWatch) with automated retraining triggers, maintaining accuracy within 2% of baseline across 6+ months.

Tata Consultancy Services (TCS) · Mar 2022 – Sep 2025

Architected GPT-3.5-Turbo audio audit framework automating QA across 50,000+ monthly call recordings eliminated 8 FTE-hours/day of manual review, reducing audit cost by 42%.
Built production multi-document RAG (LangChain + OpenSearch) using RAG Fusion, Hybrid Search, and Contextual Compression, achieving 91% answer relevance (RAGAs); reduced compliance document query time from 3 days to <5 minutes (97% reduction).
Delivered real-time Risk Analytics engine (Autoencoder + XGBoost on SageMaker) to score IT change requests, reducing production outages by 34% and yielding $2M+ annual cost avoidance.
Developed LLM agentic chatbot with tool/function calling for confidential document intelligence, cutting analyst research turnaround from days to minutes and improving self-service adoption by 65%.
Built BERT-based (Hugging Face) real-time sentiment analysis on 100K+ daily ITSM incidents, surfacing dissatisfaction signals and improving proactive escalation rate by 28%.
Created CTGAN synthetic data framework (SDV) for GDPR-compliant QA datasets, reducing test data provisioning time by 70% and unblocking downstream release cycles.

SCANFII Software Pvt. Ltd. · Sep 2020 – Mar 2022

Built CNN-based reverse image search engine for 200K+ retail products (Kmart) with NLP-driven personalization, cutting product search time by 45% and lifting user engagement by 22%.
Designed RFM + K-Means segmentation on 3M+ customer records producing 8 behavioral cohorts, enabling targeted campaigns that lifted conversion rate by 17% and reduced churn by 14%.
Deployed Prophet-based time series forecasting (Azure) for HR staff engagement, achieving MAPE 8.3% and extending planning horizon from 2 to 6 weeks ahead.
Constructed collaborative filtering recommender systems on AWS EMR, generating personalized recommendations with 35% click-through improvement over rule-based baselines.

No completed sessions yet. Anything Parthu finishes shows up here with the evidence behind it.

Jawaharlal Nehru Technological University (JNTU) · 2016 — 2020