AI/ML Engineer with 4+ years building production Generative AI and Machine Learning systems across Legal Tech · FinTech · Healthcare · Industrial domains.
"My core strength is turning complex, messy business problems into working AI."
apuroop = {
"role": "AI / ML Engineer | GenAI & LLM Specialist",
"experience": "7+ years",
"location": "Atlanta, GA 🇺🇸",
"current": "KKRGenAI Innovations LLC",
"specialties": ["Agentic AI", "LLM Fine-Tuning", "RAG Pipelines",
"Multi-Agent Systems", "MLOps"],
"llms": ["GPT-4", "Gemini", "LLaMA", "Mistral", "Claude"],
"fine_tuning": ["LoRA", "QLoRA", "PEFT", "RLAIF", "OpenAI FT API"],
"cloud": ["GCP", "AWS", "Azure"],
"superpower": "Ship AI that actually works in production 🚀",
}| 🤖 Agentic AI | 🧠 LLM Fine-Tuning | 🔍 RAG & Vector Search |
|---|---|---|
| LangGraph stateful agents | LoRA / QLoRA on Mistral-7B & LLaMA | HyDE · Parent-Doc Retrieval |
| CrewAI multi-agent systems | PEFT · RLAIF · OpenAI FT API | Pinecone · ChromaDB · Weaviate |
| Tool-calling · ReAct loops | Ada-002 · Custom embeddings | Semantic chunking strategies |
| ☁️ Cloud & MLOps | 📊 Data Engineering | 🖥️ NLP & CV |
|---|---|---|
| GCP · AWS · Azure | PySpark · Kafka · Airflow | Hugging Face · spaCy · NLTK |
| Docker · Kubernetes | BigQuery · Snowflake · dbt | NER · Summarization · Sentiment |
| MLflow · DVC · CI/CD | ETL/ELT Pipelines | OpenCV · YOLO · CNNs |
| 📋 Project Management | 📈 Business Analysis | 🤝 Collaboration & Productivity AI |
|---|---|---|
| Notion AI · ClickUp AI | Requirements & User Stories | Confluence · Miro · Slack |
| Jira · Agile / Scrum | Stakeholder Management | n8n Automation · Zapier |
| Roadmap Planning | Process Optimisation | Microsoft Copilot · ChatGPT Teams |
🏢 KKRGenAI Innovations LLC, USA | AI / ML Engineer | Jan 2025 – Present
- 🔷 Built multi-agent systems (LangGraph + CrewAI) for automated contract analysis, clause extraction, and compliance checks — in production
- 🔷 RAG pipelines over millions of legal documents using LangChain + Pinecone (HyDE, parent-document retrieval)
- 🔷 LoRA / QLoRA / PEFT fine-tuning on LLMs for legal domain adaptation — reduced hallucinations on jurisdiction-specific terminology
- 🔷 GCP-primary deployment (Vertex AI, BigQuery, Cloud Functions) + AWS SageMaker + Azure
- 🔷 n8n automation saved team 10+ hrs/week · Prompt compression & caching cut inference costs by ~30%
- 🔷 FastAPI microservices + Streamlit dashboards for legal teams
GPT-4 Gemini LLaMA LangGraph CrewAI LangChain Pinecone GCP AWS Docker LoRA QLoRA
🏢 Pragma Info Systems, USA | Data Scientist / AI-ML Engineer | Sept 2023 – Dec 2024
- 🔷 Pharmacovigilance NLP — detecting adverse event signals from clinical narrative text at scale
- 🔷 Multi-cloud workloads across AWS Bedrock, Azure, GCP · Dockerized GPU batch jobs
- 🔷 RLAIF fine-tuning + semantic retrievers using SLMs, VLLMs, Ada-002 + custom embeddings
- 🔷 PyOD signal detection — flagged drug-event associations missed by rule-based approaches
- 🔷 RAG on ChromaDB + Pinecone · FastAPI endpoints · Pydantic schema validation
OpenAI AWS Bedrock Azure ChromaDB Pinecone LangChain FastAPI PyOD Pydantic
🏢 Intellectuals AI, India | Software Developer – ML | Feb 2020 – Feb 2022
- 🔷 NLP analytics pipelines using Knowledge Graph, Ontology, WMD, NLTK
- 🔷 Web scraping (BeautifulSoup, Selenium, Scrapy) + ETL into MongoDB
- 🔷 CNN/ANN models in Keras/TensorFlow for computer vision and information extraction
Python NLTK TensorFlow Keras MongoDB Selenium Scrapy
|
|
|
|
