AI Engineer
& Backend Developer

Architecting secure, production-grade GenAI and LLM applications that drive scalable efficiency in regulated domains.

MY WORK

About Me

I am an AI Engineer with 3 years of experience specializing in secure RAG architectures, AI Agents,LLM fine-tuning, and scalable API-based AI services.

My work focuses strongly on reliability, auditability, and cost-efficient AI deployment. I have successfully delivered an 80% reduction in manual processing across workflows while maintaining 99.9% uptime on AWS.

3+ Years Experience
80% Manual Effort Reduction

Capabilities

The technical engine powering scalable and secure AI architectures.

01

Generative AI & LLMs

Llama, JAIS, Qwen, GPT, Claude, Gemini, BERT, Hugging Face Transformers, vLLM, PEFT (LoRA, QLoRA).

02

Retrieval & Agentic Systems

RAG, LangChain, LangGraph, Hybrid Search (Dense + BM25), RRF, Cross-Encoder Reranking, HNSW Indexing, Text-to-SQL, RAGAS.

03

Backend, APIs & Databases

Python, SQL, FastAPI, WebSockets, Pydantic, PostgreSQL, DynamoDB, Redis, ChromaDB, Vespa.

04

MLOps & Cloud Infrastructure

AWS, Docker, Kubernetes, MLFlow, Weights & Biases, CI/CD, Model Monitoring, Drift Detection.

My Experience

Alphastream.ai

AI Engineer — NLP

Jun 2023 — Dec 2025
  • Designed and deployed scalable NLP pipelines for financial and legal document automation on AWS, processing hundreds of documents daily with 99.9% production uptime.
  • Reduced manual document review effort by 80% (financial workflows) and 60% (legal contracts) through fine-tuned classification and structured data extraction systems.
  • Fine-tuned BERT and LegalBERT models using PyTorch and Hugging Face PEFT (LoRA), improving multi-label classification F1 score from 0.70 to 0.85+ via data augmentation and hyperparameter optimization.
  • Architected Hybrid RAG pipelines (Dense Embeddings + BM25 + RRF + Cross-Encoder reranking) enabling high recall multi-hop document intelligence for regulated workflows.
  • Built session-aware agentic systems using Redis and DynamoDB to enable persistent multi-turn AI workflows for legal analysis applications.
  • Prepared COCO-format datasets and trained YOLOv8 models for table structure detection (rows, columns, headers) to automate structured financial data extraction.
  • Collaborated with product, compliance, and engineering teams to deliver auditable, regulation-aligned AI architectures suitable for enterprise deployment.