Headline

AI Engineer (GenAI)/Data Scientist@Straive | ML and Deep Learning enthusiast | Ex-Research Intern (IISc & IITR) | ECE’24 @NITRR(Honors)

About

I’m an 𝐀𝐬𝐬𝐨𝐜𝐢𝐚𝐭𝐞 𝐃𝐚𝐭𝐚 𝐒𝐜𝐢𝐞𝐧𝐭𝐢𝐬𝐭/𝐆𝐞𝐧𝐀𝐈 𝐄𝐧𝐠𝐢𝐧𝐞𝐞𝐫 at Gramener (a Straive company) passionate about Machine Learning, Deep Learning and Generative AI. I’ve worked on building 𝐑𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥-𝐀𝐮𝐠𝐦𝐞𝐧𝐭𝐞𝐝 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 (𝐑𝐀𝐆) system for contextual data retrieval and am currently focused on 𝐚𝐮𝐭𝐨𝐦𝐚𝐭𝐢𝐧𝐠 𝐚 𝐡𝐲𝐩𝐨𝐭𝐡𝐞𝐬𝐢𝐬 𝐠𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐚𝐧𝐝 𝐭𝐞𝐬𝐭𝐢𝐧𝐠 𝐩𝐢𝐩𝐞𝐥𝐢𝐧𝐞 to accelerate data-driven decision-making. I enjoy exploring cutting-edge AI tools and developer environments, from 𝐂𝐮𝐫𝐬𝐨𝐫 and 𝐖𝐢𝐧𝐝𝐬𝐮𝐫𝐟 to emerging LLM orchestration frameworks and continuously experiment with ways to make intelligent systems more explainable and efficient. My research paper, “Deep Ensemble Transfer Learning Framework for Enhanced Endoscopic Image Classification,” was accepted at ThinkAI 2024, achieving 𝟗𝟖% 𝐚𝐜𝐜𝐮𝐫𝐚𝐜𝐲 and outperforming existing models in medical image classification. During my time in college (𝐍𝐈𝐓 𝐑𝐚𝐢𝐩𝐮𝐫), I completed 𝐭𝐰𝐨 𝐫𝐞𝐬𝐞𝐚𝐫𝐜𝐡 𝐢𝐧𝐭𝐞𝐫𝐧𝐬𝐡𝐢𝐩𝐬 at 𝐈𝐈𝐒𝐜 𝐁𝐚𝐧𝐠𝐥𝐨𝐫𝐞 and 𝐈𝐈𝐓 𝐑𝐨𝐨𝐫𝐤𝐞𝐞 in machine learning, deep learning, and computer vision at , gaining strong practical experience in model training in different fields ranging from neuroscience to time series. At Gramener, I was honored with the 𝐑𝐢𝐬𝐢𝐧𝐠 𝐒𝐭𝐚𝐫 𝐀𝐰𝐚𝐫𝐝 for delivering impactful and scalable AI solutions. I love designing systems that merge intelligent automation with real world usability, combining robust backend APIs with intuitive interfaces. 𝐓𝐞𝐜𝐡 𝐒𝐭𝐚𝐜𝐤: Python • LLMs • LangChain • FastAPI • JavaScript • DBMS • SQL • Cloud • Git • Prompt Engineering • RAG Systems • GenAI Toolchains

Location

Chennai

AI stack

anthropic apis

langchain

mcp

opencode

claudecode

RunPod

Work experience

AI Engineer at Straive, Sep 2024 - Present · 2 y, Chennai, Investment Intelligence & Actionable Insights Engine - Architected an event-driven investment platform using Azure Service Bus and Azure Functions, cutting manual investor analysis from weeks to 25 minutes while achieving 90% signal accuracy. - Built a hybrid extraction pipeline combining Anthropic Claude Web Search with a deterministic validation layer (URL verification, exact-quote matching, domain shortlisting) to eliminate hallucinations and bypass unnecessary LLM calls. - Constructed a Microsoft Fabric ETL pipeline adhering to Medallion Architecture (Bronze, Silver, Gold) to fuse verified public signals with internal CRM data and Market data which in turn are used to generate Actionable Insights. - Developed an agentic assistant using the Claude Agent SDK enabling text-to-DAX generation against Fabric Semantic Model with built-in cost tracking, evaluation benchmarks, and latency optimizations. Contextual RAG for Multi-Format Document Querying - Designed and deployed a Contextual RAG system for multi-format document querying with custom chunking and ingestion using Azure OCR, enabling extraction from PDFs, images, and text files. - Built contextual retrieval with prompt caching for Anthropic models, optimizing token usage within the context window and retaining relevant context across queries. - Implemented fusion retrieval combining cosine similarity and BM25 keyword search with pgvector embeddings in PostgreSQL. - Developed FastAPI-based microservices for ingestion, embedding, and retrieval operations, deployable across multi-cloud environments (Azure, GCP). HypoGen - Automated Hypothesis Generation and Testing Platform - Architected a modular statistical testing platform that automates hypothesis discovery, validation, and busi- ness insight generation on enterprise datasets. - Developed FastAPI back-end microservices supporting custom hypothesis modes (random, metric-focused, user-defined) and a LangChain conversational data interface. - Created an interactive JavaScript visualization dashboard equipped with automated PDF reporting to trans- late statistical outcomes into actionable campaign strategies. Multimodal LLM - OCR Pipeline for Automated Complex Table Extraction from Documents - Built an automated document extraction engine utilizing Azure OCR, LlamaParse, PyMuPDF, and Mistral AI to parse complex heterogeneous tables into Markdown and JSON. - Conducted benchmark evaluations comparing OCR-only against OCR+LLM hybrid techniques, reaching >90% accuracy on simple tables and 75–80% accuracy on multi-page structures. - Designed downstream transformation workflows to clean and standardize layout extractions for direct con- sumption by analytics pipelines. Structured Data Matching Engine - Developed a scalable data matching engine achieving 88% accuracy by blending LLM semantic analysis, fuzzy matching, and vector-driven retrieval. - Transitioned initial LLM-only matching to native PostgreSQL matching algorithms and embedding pipelines, optimizing system latency and operational costs. - Integrated Gemini API web grounding to enrich incomplete records with real-time web metadata prior to final matching. Enterprise AI Solutioning & Rapid Prototyping (POCs & RFPs) - Architected and delivered high-impact AI/ML Proof of Concepts (POCs) and technical proposals (RFPs) across Oil & Gas, E-Commerce, Investment Management, and Financial Services. - Built domain-specific document reconciliation engines, automated anomaly-detection pipelines, and agentic workflows tailored to complex, multi-source enterprise data environments. - Accelerated pre-sales engineering and client acquisition by translating complex business requirements into production-grade architectural blueprints, working prototypes, and executive intelligence dashboards. Tools & Technologies: Python, FastAPI, PostgreSQL, pgvector, Redis, Azure OCR, Anthropic Claude, Gemini, Mistral AI, Llamaparse, PyMuPDF, LangChain JavaScript, Prompt Engineering, Multimodal LLM Workflows, Data Analytics, Prototyping, Microsoft fabric, Agent SDK, LanGraph, Azure Functions, IDEs(Cursor, Windsurf, VS)

Summer Research Intern at Indian Institute of Science (IISc), May 2023 - Jul 2023 · 2 mo, Bengaluru, Karnataka, India, – Led data preprocessing and analysis for EMG data in the Neuroscience department. – Utilized Numpy, Pandas, Seaborn, Plotly, and Matplotlib to clean and analyze EMG data, driving actionable insights through exploratory data analysis and visualization. – Applied advanced machine learning techniques, such as PCA followed by Kmeans, and explored ICA for decom- posing noisy EMG data, aiming to identify early recruitment patterns of motor units.

Summer Research Intern at SPARK ( Summer Internship at IIT Roorkee ), IITR, May 2022 - Jul 2022 · 2 mo, Roorkee, Uttarakhand, India, – Spearheaded Data Science and Machine Learning initiatives at SPARK 2022, analyzing Delhi’s Electronic Ticketing Machine time series data. – Applied data cleaning and EDA techniques with Numpy and Pandas, unveiling insights and crafting compelling visualizations (histograms, heat maps, OD matrix) using Seaborn, Plotly, and Matplotlib. – Demonstrated proficiency in forecasting research by employing LSTM, RNN, GRU, and Darts, wherein LSTM consistently outperformed alternative models. – Assessed RNN-based models, employing MAPE scores as a metric in a rigorous trial-and-error research method- ology

Education

Bachelor of Technology - BTech, Electrical, Electronics and Communications Engineering, National Institute of Technology Raipur

Public links

github: https://github.com/Sadhvika-Chaparla