Data Scientist and ML Engineer with 5+ years of experience building production systems that turn messy, unstructured data into reliable outcomes. I work across NLP pipelines, transformer-based classifiers, recommendation systems, and agentic AI — from knowledge graphs and RAG to multi-agent LangGraph systems, coding agents, and hallucination detection. Currently based in Berlin, working on legal and financial document intelligence at Intapp.
0 Projects delivered
Focus: Explainable AI, Information Retrieval, Data Mining. Thesis: Script-Aware CNN-FPN Transformer for Offline Handwritten Urdu OCR.
Focus: Deep Learning, Machine Learning, Data Mining. Coursework included Text Retrieval, Big Data Analytics, and NLP.
Focus: Artificial Intelligence, Computational Intelligence, Pattern Recognition. Thesis: Image Segmentation using Autoencoders.
Designed a NER system using LUKE, extracting 5 business entities with 88% test accuracy. Led relation extraction for strategic business relationships, improving accuracy by 20%. Building document parsing inside a RAG pipeline for financial and legal text.
Built ChatGPT-powered mentoring systems that increased user engagement by 40%. Delivered AI-driven performance tracking that improved client satisfaction and retention by 30%.
Improved automated resume parsing, increasing clean data extraction from 2% to 13%. Conducted feature engineering that improved pairwise ranking scores by 2% for candidate shortlisting.
Developed a recommendation system that increased sales by 20% across 500+ stores. Built unsupervised analysis tools for non-technical clients, improving customer satisfaction by 35%.
Built a Graph Neural Network for brand recommendations, increasing user engagement by 15%. Proposed a BERT-based call transcription classifier that improved outreach efficiency by 25%.
Studied CNN-based AI systems and deployed neural network code on Raspberry Pi 3.
Analyzed enterprise network infrastructure and designed topology diagrams under bandwidth constraints.
Agentic systems, production NLP, and applied ML — from multi-agent research pipelines to industry recommendation engines.
Four-phase multi-agent system: LangGraph pipeline, FastAPI with multi-tenancy, an MCP server with three tools (tested via streamablehttp_client), and citation graph traversal using arXiv + Semantic Scholar.
LLM drift detection with Langfuse v4, a fully async LangGraph workflow, an agentic RAG diagnoser, and aiosqlite for persistent monitoring state.
SSE streaming agent with structured Pydantic outputs, LangSmith tracing, a 55-test eval framework, Docker sandboxing with hard security flags, and multi-tenancy via X-Tenant-ID.
Iteratively fixes its own output using subprocess feedback and ThreadPoolExecutor — closing the loop between generation, execution, and repair.
Hallucination detection grounded in CourtListener and SEC EDGAR sources — verifying LLM claims against real legal and financial documents. Deployed to Railway.
LUKE-based NER extracting 5 business entities at 88% accuracy, plus relation extraction for strategic partnerships.
Document parsing inside a RAG pipeline for legal and financial text retrieval.
Resume parsing and pairwise ranking for candidate shortlisting — clean extraction up from 2% to 13%.
Real-time product recommendations across 500+ stores with a 20% sales uplift.
Graph Neural Network for personalised brand suggestions, plus a BERT call-transcript classifier.
Script-aware CNN-FPN Transformer for offline handwritten Urdu recognition.
Articles on machine learning, data science, and career growth.
This article is for the students or beginners who want to pursue their career in AI or ML. This will be helpful for people who want to get familiar with these fancy words like deep learning or machine learning.
This article is especially for students who are motivated to study abroad for their masters and PhD studies. In every university, you need following documents as compulsory.
A beginner-friendly guide to the machine learning workflow — from understanding the problem domain to deploying a model in production.
Interested in data science, machine learning engineering, and NLP roles — especially where unstructured text meets production systems.
Based in Berlin. Open to roles in data science, ML engineering, and NLP.
Berlin, Germany