06/15/2026 updated
100 % available
AI/ML Engineer | LLM & Agentic Systems | Financial Compliance & Document Intelligence
Bengaluru, India
Only remote
Dual Degree (B.Tech + M.Tech) in Electrical Engineering - Indian Institute of Technology, BombayAbout me
AI/ML engineer (IIT Bombay, US patent in time-series forecasting). Building production agentic AI systems — autonomous LLM agents, RAG, structured extraction. 5+ yrs shipping at scale (380M+ records). Niche: predictive maintenance & anomaly detection for manufacturing/industrial AI.
AI AgentsRAGLLMData AnalysisFinancial Data AnalysisComputer VisionAutomationInformation RetrievalPython (Programming Language)Machine LearningPattern RecognitionPredictive MaintenanceProcurement ManagementStatistical EvaluationsTime Series
AI/ML Engineer specializing in agentic AI systems, industrial AI, time-series forecasting, and document/process automation. 5+ years building and shipping production ML systems end-to-end - from data pipelines to deployed agentic platforms serving real clients.
Agentic AI & LLM systems
- Designed and built end-to-end agentic platforms — multi-step autonomous workflows across tender review, BOM/SOR structuring, vendor management, quotation parsing (2-10x faster execution)
- Production LLM pipelines for structured extraction, document intelligence, and autonomous agents
- Self-hosted LLM inference (Qwen-2.5-14B, vLLM) for data residency and security requirements
- OCR + LLM extraction pipelines (PaddleOCR) for large PDFs (up to 1GB) and complex Excel data
- Text-to-SQL copilots, semantic search, RAG systems
Core specialization: Non-stationary time-series & industrial signal processing
- Regime-detection and pattern-specific forecasting (US patent co-inventor, IIT Bombay research)
- Applications: predictive maintenance, anomaly detection, sensor/IoT data, rig state detection (97%+ accuracy), signal classification (92%+ precision, 100% recall)
Financial compliance & risk systems
- LSTM-based transaction monitoring at scale (380M+ UK transactions, 3x recall improvement)
- Sanctions intelligence and KYC automation (96% extraction accuracy)
- Model explainability (SHAP) for regulatory compliance (SR 11-7)
Tech stack: Python, PyTorch, LLM APIs (OpenAI, Anthropic, open-source), agentic frameworks, vector databases, SQL, cloud deployment, SHAP, PaddleOCR
Background: IIT Bombay (Dual Degree, Electrical Engineering - Signal Processing), Swiss Mobility Scholar (University of Geneva)
Portfolio with detailed case studies: https://arbaz1000.github.io
Experience
Professional experience5 - 10 years
Freelancer since2025
Languages
EnglishFluent
Project history
Shipped an MVP for a pre-seed startup for tender-driven industrial/CapEx projects, achieving 2-10x faster execution across procurement stages with 20+ paid clients. Designed an end-to-end agentic backend system with a modular rule engine, built document-understanding pipelines with PaddleOCR and LLMs, exposed modular backend APIs for ERP integration, and built a product hierarchy of 7k+ products.
First AI/ML hire at a seed startup. Mentored 6 AI interns and 4 junior engineers. Built LSTM-based risk pipeline over 380M UK transactions with 3x lift in Recall@K. Designed model explainability layer using SHAP and LLM-based summaries for SR 11-7 compliance. Built LLM-driven sanctions intelligence feature with 96% accuracy. Deployed private LLM inference stack using self-hosted Qwen-2.5-14B with vLLM. Built additional compliance systems including KYC document extraction, address verification, merchant risk monitoring, adverse media knowledge graph pipeline, TEXT-to-SQL copilot, and semantic name search.
Built a RAG-based medical AI chatbot and explored LLM use for clinical coding at a pre-seed startup.