Headline

AI/ML Engineer @ Tiger Analytics | Agentic AI Systems • LLM Applications & Evaluation • RAG | IIT - BHU

About

I’m an AI/ML Engineer at Tiger Analytics, with 4+ years of experience building machine learning solutions, intelligent automation systems, and AI-powered products that solve real-world business problems. My work sits at the intersection of Agentic AI, LLM applications, and applied machine learning. I enjoy turning complex workflows into reliable, scalable systems - from designing retrieval pipelines and agentic workflows to building optimization engines and automated decision-support tools. WHAT I WORK ON ▸ Agentic AI & LLM Applications Building agentic workflows using LangChain, tool/function calling, and structured validation to automate complex, multi-step business processes. I developed an Agentic AI system for a Business Optimizer platform that reduced scenario creation time by 75% and eliminated 6–8 manual steps for users. ▸ RAG & Retrieval Systems Working with hybrid retrieval, combining lexical and semantic search with cross-encoder reranking to improve candidate retrieval for fine-grained product taxonomy classification. I also work on evaluation and guardrails to identify high-confidence false positives and improve system reliability. ▸ Applied Machine Learning Building ML-driven solutions across classification, forecasting, and optimization. My experience includes hierarchical classification under long-tail imbalance and noisy labels, long-horizon forecasting for CPG products, and trade promotion optimization to help businesses evaluate sales and profitability trade-offs. ▸ Production-Oriented AI Engineering Designing automated data pipelines, deterministic evaluation workflows, and validated agent interactions that make AI systems more reliable and useful in real-world settings. TECHNICAL TOOLKIT Python • PySpark • LangChain • LangGraph • RAG • LLM Inference • Tool Calling • Machine Learning I’m particularly interested in improving how LLM-powered systems retrieve information, reason through tasks, use tools, and are evaluated for quality and reliability. Beyond building AI applications, I’m keen to explore deeper questions around agent evaluation, reasoning, and AI safety, with a long-term interest in contributing to the development of more capable, reliable, and responsible AI systems. Outside work, I’m usually at the gym, experimenting with new recipes, or finding my way back to reading and dance.

Location

Bengaluru

AI stack

claude code

langchain

notebooklm

openai

CrewAI

Humanloop

Work experience

AI/ML Senior Associate at Tiger Analytics, Jan 2026 - Present · 11 mos, Bengaluru, Karnataka, India, PROJECT I | Agentic AI Product Taxonomy Classification ▸ Built a hierarchical GenAI-based product taxonomy classification system for fine-grained e-commerce categorization across 200K+ products and 10 vendors, addressing severe long-tail class imbalance and noisy labels. ▸ Designed a two-stage RAG pipeline combining hybrid retrieval (lexical and semantic search) with cross-encoder reranking to retrieve and rank relevant taxonomy candidates for product–category matching. ▸ Developed a custom Full-Text Search (FTS) guardrail to detect and penalize high-confidence hallucinations, mitigating plausible false positives. Evaluated LLM performance using Top-k Hit Rate, Exact Match Accuracy, and Micro-Average Precision. PROJECT II | Agentic AI for Business Optimization ▸ Built an Agentic AI system within a Business Optimizer platform to automate multi-scenario generation, reducing user effort by 6–8 steps and scenario creation time by 75%. ▸ Designed deterministic KPI evaluation pipelines to compare and rank generated scenarios, supporting consistent, data-driven business decision-making. ▸ Architected agent workflows using LangChain, tool/function calling, and Pydantic-based validation to enable reliable, scalable, production-oriented AI decision systems.

Senior Data Science Analyst at Tiger Analytics, Jul 2024 - Dec 2025 · 1 yr 6 mos, Bengaluru, Karnataka, India, ▸ Led a team of 3+ data scientists to design and build a long-horizon forecasting system predicting GSV for CPG products over a 3-year horizon. ▸ Developed a scalable, end-to-end automated data pipeline leveraging PySpark for backend processing, Azure DevOps for CI/CD with integrated unit and functional testing, and orchestrated data ingestion into a SQL database using Azure Data Factory

Data Science Analyst at Tiger Analytics, Apr 2023 - Jun 2024 · 1 yr 3 mos, Bengaluru, Karnataka, India, Built a Trade Promotion Optimization and Simulation platform to maximize MAC by optimizing promotion timing and duration, enabling scalable what-if analysis for business stakeholders.

Online Data Analyst at TELUS International AI Data Solutions, Mar 2022 - Apr 2023 · 1 yr 2 mos, ▸ Analysed and rated over 1000+ tasks and queries to improve efficiency of company-based AI search engine. ▸ Identified the query and the query intent based on given guidelines, researched about the query in detail and gave information and feedback on the results generated by the search engine. ▸ Researched over 15 different kinds of queries to help enhance various features of the search engine.

Data Scientist - Intern at TML Business Services Limited, Jun 2021 - Aug 2021 · 3 mos, ▸ Trained a neural network to identify numbers engraved on steel and auto-parts. ▸ Assembled the custom database from scratch with over 70 images. o Made 36 classes in the database including the letters of the alphabet and digits 0-9. o Manually annotated the images with the help of ‘LabelImg’. ▸ Employed the Image Recognition Algorithm - YOLO to train the neural network.

Education

Master's equivalent (Integrated Dual Degree), Bioengineering and Biomedical Technology, Indian Institute of Technology (Banaras Hindu University), Varanasi

HSC, Dr. Ambedkar College, Nagpur

Public links

github: https://github.com/sharvali-shinde

portfolio-rust-beta-55.vercel.app: https://portfolio-rust-beta-55.vercel.app/