Aman Jaipuria

Data Science Professional with 11+ years of experience leading cross-functional teams and delivering end-to-end Machine Learning, NLP, and Generative AI solutions across BFSI, Manufacturing, Product, and Supply Chain domains. Brings versatile, hands-on expertise across the full data spectrum - from predictive modeling, data pipelines, and Databricks engineering to process automation and advanced AI - including Agentic AI systems, MCP-based tool integrations, and vibe-coding workflows. A fast learner who rapidly adopts emerging technologies, tools and frameworks, translating them into production-ready, solution-oriented outcomes.

Proven ability to own full model lifecycles, from stakeholder alignment and roadmap planning to MLOps grade deployment on AWS, Azure and Databricks, while managing and upskilling high-performing teams. Deep expertise in Python, PySpark, SQL, LLMs, Agentic AI, and MLflow, with a track record of translating AI strategy into measurable business outcomes.

Data Science Manager at ACUITY ANALYTICS (2024-09 – Present)

• Designed and implemented Prism, a self-evolving, self-maintaining GenAI decision-intelligence platform. Prism transforms conversational telemetry into self-refining data skills and serves as a substitute for traditional business intelligence reporting. It achieves this by delivering automated SQL generation, real-time metric streams, and live key performance indicators (KPIs) across various Fixed Income datasets in BigQuery and PostgreSQL.
• Prism features a closed-loop, multi-agent learning architecture that incorporates human-in-the-loop (HITL) verification. This allows it to dynamically extract business and technical insights from user interactions, continuously improving its capabilities. Additionally, it is supported by an automated metacognitive maintenance agent that audits skill integrity, translates findings into skill updates, establishes new data topics, and automatically logs engineering tasks in Jira to prevent performance drift.
• Designed, developed, and deployed a machine learning-based Client Recommendation System for Axed Bonds, achieving 84.5% model accuracy and surfacing the top 15 buyers/sellers per bond; actively used by 50+ sales users alongside a Client Profiling module, directly accelerating bond sales decisions.
• Architected an end-to-end trading data pipeline on Databricks, powering the All Trades Insights dashboard used by 25+ front-office traders for real-time, data-driven decisions.
• Led and managed a team of 3, looked after project allocation, team deliverables, and sprint planning using Agile/Kanban, ensuring on-time delivery across concurrent data science projects.
• Conducted technical interviews for Data Scientist roles (associate/senior-associate/lead levels), evaluating candidates on ML, Python, SQL, LLMs, and system design.
• Active contributor to the ML Community of Practice, sharing hands-on knowledge of Agentic AI, MCP, and the latest GenAI advancements to drive continuous upskilling across data science teams.

Data Science Lead at CUMMINS (2023-09 – 2024-08)

• Built a Generative AI tool using Azure OpenAI and NLP to auto-extract insights from 10-K reports, reducing research time from days of reading 100+ pages to minutes and enabling sales teams to identify high-potential targets faster.
• Developed a RAG-based chatbot PoC using LLMs and LangChain, subsequently integrated as a core feature of the 10-K GenAI tool for on-demand interactive querying.
• Built an Intelligent Summarization tool using pattern matching and NLP to highlight relevant codes and sentences in customer claim narrations, resulting in 20–25% effort savings per claim.
• Led and managed a team of 4, overseeing task allocation, project delivery, and day-to-day technical guidance across concurrent client engagements.

Data Science Specialist at AVERY DENNISON (2021-05 – 2023-08)

• Deployed two ML models as Flask API endpoints: a Claim Disposition Duration predictor and an Auto-Approval Recommendation model validated for phased production deployment to streamline low-risk claim processing.
• Designed an AI solution for the Woven Label North Asia hub to predict correct coloured yarn sets, reducing cycle time by 30% and improving dye yarn hit-rate accuracy from 57.9% to 79.7%.
• Built the Safe-Width 2.0 solution using mathematical models to recommend roll sizes with a high probability of sale, optimizing inventory and reducing cutting scrap costs.
• Applied Computer Vision, image annotation, and transfer learning to build a pallet packaging quality prediction model, increasing customer satisfaction and reducing product returns.
• Utilized RFM Analysis and feature engineering to build a Next Order Date prediction classification model for the supply chain team.
• Organized and delivered 2 Data Science bootcamps with 200+ attendees, building org-wide data literacy among non-technical employees.

Data Scientist at COSMO FILMS (2020-09 – 2021-05)

• Performed topic modeling on Google Play app reviews to provide machine learning-based vendor selection recommendations, saving an estimated 2–3 weeks of manual evaluation.
• Built demand forecasting models for the top two product categories with 85–90% accuracy.

Data Scientist at HINDALCO INDUSTRIES LIMITED (2015-07 – 2020-08)

• Applied ML prediction models on IoT and quality data to optimize process parameters, reducing wire breaks to 0.4 breaks/MT during copper rod drawing.
• Built a real-time anomaly detection solution for 12-stand rolling mills, iteratively retrained to reduce false positives and achieve high-confidence alerts that reliably indicated actual service-impacting incidents.
• Conducted ad-hoc analysis on process parameters to strengthen the copper rod coil forming process, preventing coil slippage and breakage during plant startups.

Post Graduation in Data Science – International Institute of Information Technology, Bangalore (2019 – 2020)

Bachelor of Technology in Mechanical Engineering – Indian Institute of Technology (BHU), Varanasi (2011 – 2015)