Gopi Krishna Mahankali | Data Scientist

Hi, I'm Gopi Krishna Mahankali.

A

Self-driven, quick starter, passionate programmer with a curious mind who enjoys solving a complex and challenging real-world problems.

About

I am a Electrical Engineering Grad Student at IIT Bombay. I enjoy problem-solving and coding. Always strive to bring 100% to the work I do. I have worked on various data science projects and research spans industrial AI (condition monitoring with FFT/LSTMs, >90% accuracy), causal fault diagnosis (PC/SVAR), and GenAI (RAG chatbots, Digital Twin LLM assistant). Additional projects include knowledge graph chatbots for medical data and NLP email classification systems. I have 32 months of professional work experience which helped me strengthen my experience in Python, Data Science, GenAI. I am passionate about developing complex applications that solve real-world problems impacting millions of users. I am a winner and runner-up in 3 hackathons conducted in BGSW.

Frameworks & Tools: PyTorch, scikit-learn, Docker, Azure, Git, LangChain, SQL, Transformers
Knowledge & Skills: Python, Machine Learning, Natural Language Processing, Signal Processing, Deep Learning, Generative AI, Time Series Analysis, Predictive Modeling

Looking for an opportunity to work in a challenging position combining my skills in Data Science, which provides professional development, interesting experiences and personal growth.

Experience

Bosch

Data Scientist

GROW: Digital Twin Platform
- Led a team of five in a condition monitoring project for 56+ turbomachinery assets (compressors, gas turbines, heat exchangers, pumps), employing physics-driven signal processing (FFT, wavelets, kurtogram) and machine learning models. Achieved >90% accuracy in real-time condition monitoring for clients like JSW and ADNOC, reducing downtime costs by diagnosing structural fatigue and performance issues proactively.
- Researched causal graph-based fault diagnosis systems leveraging PC, SVAR and RCD algorithms, using sensor data and domain knowledge to identify root causes and optimize turbomachinery maintenance strategies.
- Trained an LSTM Autoencoder for unsupervised anomaly detection in industrial machinery, leveraging sensor data to model healthy operational patterns. The system identifies anomalies by computing reconstruction errors and furthermore using temporal aggregated RE to generate RUL degradation model.
- Implemented an LLM-powered chatbot within the Digital Twin, enabling function calling to backend APIs for asset health insights and vector database queries for fault remedies, delivering comprehensive, context-aware responses.
Knowledge Nexus
- Built a Generative AI RAG chatbot using Mistral with a custom section-based chunking pipeline and Qdrant as the vector database for embeddings to process PDFs and Docupedia content.
- Designed a hybrid retrieval system combining BM25, BGE, and ColBERT for optimal balance between exact term matching and semantic search, with cross-encoder re-ranking.
- Integrated agentic workflows via Hugging Face Agents, equipping the chatbot with tools like internet search, retrieval, table of contents navigation, and query transformation.
POC's
- Designed and implemented a knowledge graph-based chatbot using LangChain and Neo4j, leveraging structured hospital data with LLMs to deliver accurate, hallucination-free responses.
- Improved email processing efficiency and accuracy by developing an ML-based solution using NLP techniques like TF-IDF and CBOW. Delivered a high-performing model that met client expectations for accurate email classification and routing.