Ishan Suryan

Leading the design and delivery of production-grade Generative AI solutions for enterprise clients, with a focus on RAG architectures, LLM orchestration, and conversational AI. 🔹 Sustainability Expert Chatbot (6 months) Built a domain-specific RAG chatbot grounded in Accenture's Green Initiatives and 360 Value Report. Designed a dual-pipeline architecture — an ingestion pipeline that auto-detects document changes, and a search pipeline combining vector retrieval with LLM-based chunk validation across 3 ranked knowledge sources. → 85% answer relevance | 40% fewer irrelevant responses | ~90% reduction in search times 🔹 Competitor Benchmarking Platform (6 months) Led a team of 4 to build an automated web platform that extracts and maps sustainability metrics from competitor annual reports against 52 World Economic Forum indicators — enabling structured comparison with Accenture's own declarations. → 52 metrics automated | ~80% reduction in manual benchmarking time 🔹 Dabur – AI-Powered KPI Dashboard (Ongoing) Leading a team of 5 to build an enterprise BI dashboard with a conversational AI layer for client Dabur. Designed a 2-stage backend: vector similarity search → table metadata → LLM-generated SQL → narrative insights and dynamic chart generation. → Enables natural language querying of KPIs for non-technical stakeholders | Target: sub-5s response

Applied ML and NLP solutions in an automotive manufacturing context, working on-site with Stellantis India. • Applied NLP to employee survey data to surface workforce sentiment patterns and inform HR strategy — ~88% sentiment accuracy • Developed an energy usage prediction POC for a manufacturing plant, enabling cost-optimised procurement decisions — ~12% cost reduction • Built an analytics pipeline for ADAS L2+ test vehicle inhibit/suppression event analysis, improving test cycle efficiency

Delivered Data Science & ML certification training to 100+ students across CampK12 and Webshala.in. Designed curriculum covering Python, ML algorithms, NLP, and model deployment. Coached students through campus placement preparation.

Dual role spanning AI/ML R&D and automotive engineering, with a focus on driver behaviour and vehicle intelligence. • Led the Hagora Pulse R&D project — detecting driver mental state via speech-to-text analysis using ML/DL algorithms • Built driving pattern prediction models to classify real-time driver style from sensor data • Predicted office arrival times and effective working hours using historical data regression models

Permanently placed at client site (Ford India Pvt Limited) as Senior Design Engineer Have been on site in Ford North America for 3 months for “Autonomous car training on ML and AI using Python”. •Based on car driving conditions of the driver, autonomous detection of Car Mileage using AI. •Employee Attrition prediction: Predict the employee attrition probability based on various conditions/parameters. •Car price Prediction: predict the car price based on car features. •Spam Email classifier (using NLP): classify an e-mail as ham or spam based on word usage.

• Concept realization in CADD i.e. from drawings, renderings, scan data, existing surface models and clay models. • Reverse engineering (esp. from Scan/Point cloud data), Surfacing & Solid modeling (in ICEM SURF /CATIA) for prototyping of concepts, development existing physical parts based on customer requirements. • Component Designing. • Concept generation.

• Class A Surfacing / Solid modeling (in CATIA) of proto type to concepts/drgs./physical parts based on customer requirements. • Class A Surfacing using Scan/Point cloud data input. • Concept generation. • Detailed 2D drgs. As per the 3D model.

• Class-A Surfacing and 3D modeling in ICEM SURF, Catia V5 • Class A surfacing from Scan/Point Cloud data. • Mould Design. • 2D to 3D Modeling • 3D Model to 2D Drawings • Elementary designing of Jigs and Fixtures

Currently pursuing the course. While working on several guided projects.

Successfully completed the summer training on data science while working on multiple projects.