Ayush Agrawal — AI Engineer | LLM Systems, Agentic Workflows, RAG Pipelines, Voice AI & NLP

AI Engineering Experience & Enterprise LLM Systems

Building production-grade AI systems: agentic workflows, RAG pipelines, voice AI, and document intelligence platforms deployed at enterprise scale.

ICICI Prudential AMC

AI Engineer · Contracted via Blue Altair · Full-time

JUN 2025 — PRESENT

Automated end-to-end monthly regeneration of digitalfactsheet.icicipruamc.com — eliminated a full manual rebuild every month through AI-driven pipeline orchestration and LLM-based content generation
Architected agentic n8n workflows processing 450+ investor communication mailers at scale, significantly reducing turnaround time
Built Compliance Assist & HR Assist — NLP-driven document intelligence systems across 1,500+ SEBI/AMFI regulatory documents and 200+ HR guidelines; semantic search and compliance validation
Engineered AI voice bot for SIP renewals using LiveKit, ElevenLabs (TTS), Sarvam AI (STT), and Gemini LLM — fully automated outbound investor communication
Built AI-powered marketing portal for automated landing page generation using LangChain, RAG pipelines, and Azure OpenAI
Onboarded and integrated enterprise AI platforms: ElevenLabs, Sarvam Dubbing Studio, Google Flow, and Copilot Studio — reducing vendor dependency and operational costs

n8n LangChain RAG LiveKit ElevenLabs Sarvam AI Gemini Azure OpenAI FastAPI Vector DB Copilot Studio

FactSet

AI/ML Engineering Intern

MAY 2024 — JUL 2024

Developed NLP pipeline (LangChain + BERT) to analyze 500+ meeting transcripts — automated extraction of key themes, action items, and sentiment trends, reducing manual review by ~70%
Applied GPT-3.5 for topic modeling and key business insight extraction; sentiment analysis across all transcripts
Built interactive dashboards in Tableau and Power BI surfacing actionable insights for senior stakeholders

LangChain BERT GPT-3.5 Tableau Power BI Python

Queen Mary University of London

Data Science Research Intern · Remote

JUN 2023 — AUG 2023

Applied LDA topic modeling on hydrogen energy research datasets to identify emerging research trends
Implemented TF-IDF term ranking and coherence testing; delivered findings through academic-style reports to faculty

LDA NLP Topic Modeling Python NLTK Pandas

AI & Machine Learning Projects: LLM, NLP & Voice Systems

Production-grade systems with measurable outcomes: semantic search, text analysis, and voice automation platforms.

Mutual Fund NLP Chatbot

BERT-based intent classification and NER system for financial queries. Microservices architecture via FastAPI with custom domain-specific training data and JWT authentication.

85% classification accuracy

BERT FastAPI RAG FAISS MySQL JWT

FeedbackAI

NLP pipeline converting raw user feedback into structured JSON insights using zero-shot classification and SentenceTransformers. Full analytics dashboard with sentiment, category, and aspect-level breakdowns.

10,000+ entries processed in <60s

Hugging Face FastAPI spaCy MongoDB Chart.js

NIFTY50 Backtest Engine

Backtrader-based backtesting system for MA Crossover and Bollinger Band strategies over 10 years of NIFTY50 data. Parameter optimization grid search with full PnL and drawdown analytics.

Sharpe ratio 1.8 · +12% vs buy-and-hold

Python Backtrader Pandas Matplotlib yFinance

Enterprise AI Platforms: Voice AI, LLM & Cloud Deployment

End-to-end deployment and cloud integration of enterprise AI platforms, voice systems, and orchestration workflows at ICICI Prudential AMC.

ElevenLabs

Text-to-speech for voice bot; high-quality investor-facing audio generation

Sarvam AI

Hindi/English speech-to-text for SIP renewal voice bot; regional language support

Sarvam Dubbing Studio

AI-powered multilingual dubbing for investor communication content

Copilot Studio

Internal automation bots for Microsoft Teams; cross-team workflow automation

Google Flow

Gemini-powered workflow orchestration for marketing content generation

LiveKit

Real-time audio/video transport layer for the AI voice bot infrastructure

n8n

Agentic workflow orchestration; 450+ investor mailer pipelines automated

Azure OpenAI

Enterprise GPT-4 deployment for compliant, private LLM inference at scale

LLM & AI Engineering Capabilities: Multi-Agent Systems & Voice AI

End-to-end AI engineering across the full production lifecycle: agentic workflows, semantic search, NLP systems, and enterprise voice automation.

01 / AGENTIC AI

Multi-Agent Orchestration

Designing autonomous agent systems using LangChain, n8n, and custom orchestration layers. Agents that reason, plan, and execute complex multi-step business logic without human intervention.

02 / RAG SYSTEMS

Retrieval-Augmented Generation

Building proprietary knowledge retrieval pipelines with vector databases (FAISS, pgvector), semantic chunking strategies, and re-ranking layers for high-accuracy, grounded LLM outputs.

03 / VOICE AI

Conversational Voice AI Systems

End-to-end voice AI pipelines integrating STT (Sarvam AI), LLM reasoning (Gemini), and TTS (ElevenLabs) via LiveKit transport. Real-time, low-latency systems deployed for enterprise outbound communication.

04 / NLP ENGINEERING

Document Intelligence & NLP

BERT-based NER, intent classification, sentiment analysis, and topic modeling at scale. Production NLP systems for compliance document processing, feedback analysis, and financial query parsing.

Tech Stack: LLM Frameworks, Cloud Platforms & AI Tools

LLM / AI Models

Gemini Pro GPT-4o Azure OpenAI Claude 3.5 Llama 3 BERT

Frameworks

LangChain FastAPI PyTorch Hugging Face n8n LiveKit spaCy

Cloud & DevOps

AWS Azure Google Cloud Docker Git

Databases

PostgreSQL MongoDB FAISS pgvector MySQL ChromaDB

Certifications

AWS Certified Cloud Practitioner

Amazon Web Services

Azure Fundamentals AZ-900

Microsoft

Transformers & NLP Specialization

DeepLearning.AI

Text Embeddings

DeepLearning.AI · May 2024