Rishabh Shah
AI Engineer | LLM Systems, RLHF, Evaluation Pipelines | Experience on Gemini (via Turing) | IIT Madras
Mumbai
4+ years of work experience
Dec 20256 moJun 2026
Mumbai, Maharashtra, India
May 20241 y 7 moDec 2025
Mumbai, Maharashtra, India
Worked on Google’s Gemini project (via Turing), building large-scale LLM systems focused on reasoning, reliability, and evaluation.
• Built agentic workflows and evaluation pipelines to assess LLM reasoning and robustness across diverse tasks
• Applied SFT and RLHF techniques to improve model performance, increasing accuracy from 12% to 44% on curated benchmarks
• Designed simulation environments for API-driven and multimodal tasks, enabling scalable and reproducible testing
• Developed quantitative evaluation frameworks for model consistency, edge-case behavior, and failure analysis
• Co-led cross-functional efforts across 15+ engineers to deliver ML modules and evaluation systems at scale
Jan 20245 mosMay 2024
Bengaluru, Karnataka, India
Optimized Provider Directory Accuracy and Attribution Outlier Detection systems, focusing on explainable time series modeling and scalable production efficiency.
* Built anomaly detection models (SARIMA, ARIMA, LSTM) to monitor shifts in provider attribution; adopted SARIMA for its interpretability.
* Improved entity-resolution by using DOM-based proximity logic, enhancing accuracy of provider-linkage mapping.
* Migrated modules to PySpark, cutting execution time and reducing AWS compute costs by 20%.
* Deployed ML pipelines with monitoring and explainability safeguards for production readiness.
Jul 20221 yr 6 mosDec 2023
Pune District, Maharashtra, India
Developed ML and NLP modules for HiLabs’ Provider Directory Accuracy platform — improving data integrity and automated entity resolution across large-scale provider datasets.
* Built ARIMA, SARIMA, and neural network models to detect attribution anomalies, reducing false positives by 30%.
* Applied spaCy NER and proximity-based algorithms to extract and link provider attributes from unstructured web pages.
* Created a multi-class XGBoost model using BERT embeddings, achieving 93% precision for website classification.
* Designed structured schemas linking providers to attributes like address and specialty for improved validation.
May 202111 moApr 2022
Chennai, Tamil Nadu, India
As a part of my Master's Thesis, collaborated with Saint-Gobain Research India and RBCDSAI, IIT Madras to enhance induction furnace performance through predictive modeling and optimization. The project applied industrial data science techniques to improve process efficiency and operational reliability.
* Built XGBoost, PORT, and Neural Network models on high-frequency sensor data to predict optimal operating parameters.
* Developed a Genetic Algorithm–based optimization framework, improving solver speed by 3× and enhancing interpretability.
* Implemented time series and regression models for performance forecasting and real-time monitoring.
* Bridged academic and industrial collaboration by translating research prototypes into deployable optimization workflows.
May 20213 mosJul 2021
Rishabh’s AI Stack
Education
Inter Disciplinary Dual Degree (IDDD), Data Science
Indian Institute of Technology, Madras
20175 yrs2022
Bachelor of Technology - BTech, Naval Architecture and Ocean Engineering
Indian Institute of Technology, Madras
20175 yrs2022
Join Rishabh and 10k+ other professionals on
Tal is the app where bosses meet engineers & builders directly, without middlemen
Download Tal Boss