Portfolio · 2026 · Agentic AI
Nagesh CH
AI Architect / Engineer · 11+ Years
Open to Global Opportunities
Data Science & AI/ML Architect — 11+ Years

NAGESH CH

Agentic AI Specialist · Enterprise AI Strategy

⚡ Agentic AI 🤖 LLM Engineering 📊 Data Engineering ☁️ Cloud Platforms 🏢 Enterprise AI Strategy 🔬 MLOps · LLMOps

I build production-grade GenAI, Agentic AI, and ML systems that transform workflows into trusted business outcomes. My work is designed to earn adoption, reduce risk, and move teams from proof-of-concept to production at scale.

View My Work → Get in Touch
Nagesh CH Hyderabad · Open Globally
$9M+
Business Impact
11+
Years XP
$1.5M
Largest Deal
93%
BERT Accuracy
90%
Time Reduction · askchemille.com
$1.5M Closed Sales · askchemille.com $3M Google Ad Revenue Identified 93% BERT Accuracy on 1.5M Transcripts 90% Time Reduction · Polymer Selection Zero Hallucinations in Production 60% MTTR Reduction · Sitel Telecom $9M+ Business Impact Across Engagements $1.5M Closed Sales · askchemille.com $3M Google Ad Revenue Identified 93% BERT Accuracy on 1.5M Transcripts 90% Time Reduction · Polymer Selection Zero Hallucinations in Production 60% MTTR Reduction · Sitel Telecom $9M+ Business Impact Across Engagements
Enterprises Empowered by My Solutions
Google
Waymo
Celanese
Bose
Under Armour
Kraft Heinz
ITC
TripAdvisor
Ecolab
Zenni
MCC
01
01

My Journey

"I architect AI systems that don't just impress in demos — they ship, scale, and drive real revenue."

My journey began when I saw teams struggle with data they could not trust and decisions that still relied on manual judgment. I moved from analytics dashboards to AI because I wanted systems to help people act faster with confidence.
Why I build AI systems: to turn complexity into decision intelligence, to make knowledge repeatable, and to create tools that accelerate growth without adding risk.
What matters most: not great demo metrics, but products that earn adoption, demonstrate ROI, and become a dependable part of a team’s daily workflow.
  • Enterprise copilots & autonomous chatbots
  • Multi-agent orchestration & tool-augmented agents
  • AI-driven knowledge management & intelligent search
  • GenAI-powered business automation & decision intelligence
  • Text-to-image, code, and multimodal GenAI pipelines
  • Summarization, semantic search, and compliance monitoring
Location
Hyderabad, Telangana
Open to Global Opportunities
Email
ncheripally@outlook.com
LinkedIn
linkedin.com/in/ncheripally9
GitHub
github.com/ncheripally/nagesh-ch
Certifications
ADaSci Agentic AI Architect / EngineereCornell Data Analytics 360Google PM Certificate
Awards
Pragathi Award — WiproI Made a Difference — Sitel
02
02

Work History

CELANESEJan 2025 – PresentChemical / Materials
Sr. AI Engineer / Architect

Engineered askchemille.com — a production multi-agent AI platform for polymer grade selection powering the global sales team. This case study began as a small pilot and grew into a trusted, revenue-generating sales assistant with live performance guarantees.

  • Contributed to $1.5M closed sales via AI-driven grade recommendation
  • Improved accuracy by 26% — GPT-4.1 + LLaMA-3-8B hybrid
  • Reduced comparison time by 90% using o1-mini & o3-mini
  • Cut query resolution by 40%; satisfaction up 15%
  • Orchestrator, guardrail, validator & assumptions agents in prod
  • Zero hallucinations in production deployment
LangGraphAzure OpenAIRAGGPT-4.1LLaMA-3NLP-NERAzure DevOpsCI/CD
ACIES GLOBALApr 2022 – Aug 2024Enterprise Consulting
Senior Data Scientist

Led GenAI and data science engagements for Google, Waymo, Bose, Under Armour, Kraft Heinz, and Ecolab. Managed a cross-functional team of 4 data scientists.

  • Identified $3M in Google Ad revenue via spend analysis
  • Streamlit app cut analysis time from 2 hrs → 10 min
  • RAG pipeline reduced query resolution by 40%
  • Zero hallucination/jailbreak — BERTScore + ROUGE guardrails
GPT-4oRAGBERTScoreROUGEStreamlitBigQuery
SITEL GROUPMar 2021 – Apr 2022Telecom / UCaaS
Data Scientist

Multilingual helpdesk automation, churn prediction, and intent classification at scale for a Telecom UCaaS client.

  • 60% MTTR decrease — Deep Translator NLP (5+ languages)
  • Generated $136K/year — replaced third-party tools
  • 93% BERT accuracy — 1.5M transcript classification
  • 29% retention improvement; 3.5% churn revenue reduction
BERTDeep TranslatorNLPScikit-learn
WIPROMay 2017 – Feb 2021Enterprise Analytics
Group Lead Analytics

Enterprise analytics — CLV modelling, product recommendation engines, renewal prediction, and BI dashboards for global clients.

  • RFM model on 2M customers — Gaussian Mixture + Gradient Boosting
  • 85–90% accuracy on renewal prediction models
  • Saved $224K/year — won PRAGATHI AWARD
  • Full BI stack across Power BI & Tableau for global stakeholders
Scikit-learnPySparkGradient BoostingPower BITableau
03
03

Technical Skills

LLM Engineering
LLMs & GenAI
ChatGPTGPT-4.1GPT-4oClaudeClaude 3LLaMALangChainLlamaIndexHugging FaceOpenAI APIPrompt EngineeringRAG PipelinesVector SearchEmbeddingsLoRA / QLoRAPEFTRLHF
Agentic AI
Agentic AI
LangGraphAutoGenCrewAIOpenAI Agent SDKn8nMulti-agent SystemsTool UseTask PlanningMemory ManagementRedisCognite Atlas AI
ML Frameworks
ML Frameworks
PyTorchTensorFlowScikit-learnTransformersLangFuseFastAPIFlaskMLFlow
Data Engineering
Data & Vector DBs
FAISSPineconeWeaviateQdrantBigQuerySnowflakeRedshiftSparkKafkaPandasPySpark
Cloud Platforms
Cloud & MLOps
Azure CloudAzure DevOpsDockerCI/CD PipelinesGitPostmanJIRA
Languages & Viz
Languages & Visualization
PythonSQLJavaScriptBashREST / gRPCPower BITableauPlotlyMatplotlibSeaborn
04
04

AI Demos

DEMO / 01
Multi-Agent Orchestration
Live demo of agentic AI systems with orchestrator, guardrail, and validator agents working in harmony. Full production pattern with runtime observability.
LangGraphAzure OpenAIAgentsMLOps
Request Demo →
DEMO / 02
RAG Pipeline Explorer
Interactive RAG system showing document retrieval, embedding, and LLM generation with model tracing, evaluation dashboards, and drift detection.
RAGEmbeddingsLLMModel TracingEvaluation
Request Demo →
DEMO / 03
NLP Classification Engine
BERT-based text classification and entity recognition with performance monitoring, explainability analytics, and live model validation.
BERTNLPClassificationMonitoringExplainability
Request Demo →
DEMO / 04
AI Chatbot Assistant
Conversational AI powered by GPT-4.1 and custom fine-tuning for domain-specific queries with memory, tool-use, and model evaluation workflows.
GPT-4.1ChatFine-tuningObservationModel Evaluation
Request Demo →
DEMO / 05
MLOps & Evaluation Studio
State-of-the-art lifecycle demo with CI/CD, model versioning, tracing, evaluation scoring, and governance dashboards for production AI.
MLOpsModel TracingModel EvaluationMonitoring
Request Demo →
DEMO / 06
Responsible AI Compliance
Governance-focused demo for bias mitigation, explainability, and audit trails to meet enterprise AI safety and compliance requirements.
Responsible AIBias MitigationExplainabilityAudit Trail
Request Demo →
DEMO / 07
Real-Time GenAI Monitoring
Runtime monitoring demo with live latency SLA tracking, prediction drift alerts, and feedback loop integration for production-grade GenAI services.
MonitoringDrift DetectionSLAFeedback Loop
Request Demo →
DEMO / 08
Continuous Model Delivery
End-to-end model delivery pipeline with automated retraining triggers, model versioning, and canary rollout for safe, scalable AI deployments.
CI/CDRetrainingCanary RolloutVersioning
Request Demo →

See these systems in action. Discuss which fits your roadmap.

Schedule a Demo →
05
05

Featured Projects

01 · Featured · Live in Production
askchemille.com
Celanese Corporation · Jan 2025 – Present

The Challenge: Celanese sales team spent hours comparing thousands of polymer grades manually. No system existed to recommend grades intelligently. Lost sales due to slow quote turnaround.

What We Built: Multi-agent AI platform combining custom NLP-NER for grade extraction, GPT-4.1 for reasoning, LLaMA-3-8B for cost efficiency, and RAG for technical docs. Orchestrator-validator-guardrail pattern ensures zero hallucinations in production.

The Win: Sales team now closes deals 10x faster. $1.5M in direct revenue attributed. Customers can now self-serve with confidence. Platform is mission-critical for global operations.

GPT-4.1LLaMA-3-8BLangGraphAzure OpenAIRAGNLP-NERAzure DevOps
  • Sales contribution$1.5M
  • Recommendation accuracy+26%
  • Comparison time reduction−90%
  • Query resolution time−40%
  • Customer satisfaction+15%
  • Production hallucinationsZero
02 · Growth Optimization
Google Ad Spend Optimization
Acies Global · Google Ad Analytics

Found $3M in revenue opportunities through granular bid modeling and cross-channel attribution. Built Streamlit dashboard that cut analysis from 2 hours to 10 minutes.

  • Revenue opportunity identified$3M
  • Optimization uplift+18%
  • Analysis time reduction−75%
BigQueryGoogle AdsPythonSQLStreamlit
03 · Enterprise Consulting
GenAI Product Intelligence
Acies Global · Google, Bose, Under Armour

GPT-4o review summarisation, RAG support pipeline, BERTScore + ROUGE eval guardrails — zero hallucination in production.

  • Ad revenue identified$3M
  • Analysis time (2hrs→10min)−92%
  • Query resolution−40%
GPT-4oRAGBERTScoreROUGEStreamlit
04 · Telecom / UCaaS
Sitel Group Helpdesk AI
Sitel Group · Mar 2021 – Apr 2022

Multilingual ML helpdesk system for ticket closure, churn prediction, and intent classification across 1.5M transcripts.

Delivered end-to-end production ML ownership, deprecated third-party vendor tools, and presented outcomes to stakeholders while saving cost and improving automation.
  • MTTR reduction60%
  • Annual savings$136K
  • BERT classification accuracy93%
  • Retention improvement29%
  • Churn revenue reduction3.5%
BERTDeep TranslatorNLPScikit-learnPySpark
05 · AI Governance
Production AI Governance Studio
Enterprise AI Programs · Cross-functional Teams

Built a model evaluation, tracing, and governance studio with scorecards, audit logs, and deployment safety checks for enterprise GenAI systems.

  • Risk reduction30%
  • Models traced15+
  • Review cycle reduction65%
MLFlowAzure DevOpsRAGBERTScoreGovernance
06
06

Hard Problems Solved

01
LLM hallucination in production
Implemented BERTScore + ROUGE evaluation pipeline with an automated guardrail agent. Result: zero hallucination or jailbreak incidents in Celanese's live deployment.
Guardrails · Evaluation
02
Multi-agent orchestration at scale
Designed orchestrator-validator-guardrail pattern using Azure OpenAI Agent SDK. Agents are modular and independently swappable — no cascading failures during updates.
Architecture · Resilience
03
Multilingual helpdesk at scale
Built Deep Translator + custom NLP pipeline across 5+ languages, replacing expensive third-party tools while delivering 60% faster mean time to resolution.
NLP · Cost Reduction
04
Grade recommendation accuracy
Combined GPT-4.1 with Meta LLaMA-3-8B via custom NLP-NER hybrid — 26% accuracy improvement over baseline, contributing directly to live sales deals.
Hybrid Models · Revenue
07
07

Research &Writing

Medium
AI Engineering
Building Production-Grade Agentic AI Systems with LangGraph
Deep dive into orchestrating multi-agent systems, handling fallbacks, and ensuring zero-hallucination in production.
Read →
Towards Data Science
LLM Engineering
RAG Pipelines: From Basics to Production-Ready Systems
Comprehensive guide on retrieval-augmented generation systems with evaluation metrics and guardrails.
Read →
LinkedIn
AI/ML Strategy
Enterprise AI Architecture: Scaling from POC to Production
Lessons learned deploying AI systems at scale. Best practices for MLOps, governance, and cost optimization.
Read →
Research
NLP
Fine-tuning Transformer Models for Domain-Specific NLP Tasks
Technical paper on adapting BERT and LLaMA models to specialized domains with limited data.
View →
CONNECT
Available Now · Global Opportunities

LET'S BUILDTogether

Agentic AI Specialist · Enterprise AI Strategy · Immediate availability for global roles in Singapore, Canada, Germany, UK & beyond.

Email ncheripally@outlook.com LinkedIn linkedin.com/in/ncheripally9 GitHub github.com/nagesh-ch
Start a Conversation →