AI Engineer focused on building production-grade AI systems — including LLM-powered applications, voice agents, and OCR pipelines.

I specialize in designing deterministic + hybrid AI architectures that reduce hallucination, improve reliability, and scale in real-world environments.

My work combines:

• LLMs + RAG systems
• Voice AI (STT/TTS pipelines)
• Computer Vision (OCR, detection)
• Backend systems (FastAPI, Redis, Docker)

I focus on shipping systems, not just models.

• Cygnus Payments - AI / Backend Engineer (Aug 2025 - Present)
• Independent Research - LLM Research Engineer (Jan 2025 - Present)
• Center of Language Engineering - AI Research Officer (Nov 2023 - Feb 2025)
• Nodlays - AI Engineer (Oct 2022 - Jan 2024)

Real-time AI voice ordering system using Twilio, Deepgram, FastAPI, and Redis.

• Deterministic state machine + NLU pipeline (low hallucination design)
• Real-time audio streaming with sub-second latency (<1s)
• Production-oriented architecture (session management, state routing)

CNN-LSTM based OCR + NLP pipeline for Urdu, Arabic, and Farsi.

• Achieved ~98% accuracy with CER reduction (3.4% → 2.3%)
• End-to-end pipeline: image → text → structured output
• Handles low-resource language challenges

OCR + LLM pipeline for structured menu extraction.

• Converts unstructured menus into structured data
• ~99% manual effort reduction
• Combines CV + LLM reasoning

Instruction fine-tuning with semantic batching (Flan-T5 + LoRA).

• FAISS-based grouping for stable training
• Multi-seed reproducible experiments
• Focus on convergence stability + generalization

• Designing reliable AI systems (not just prototypes)
• Reducing hallucination via deterministic pipelines
• Building scalable backend + AI integrations
• Experimentation + research (LLMs, training strategies)

AI / ML: PyTorch, TensorFlow, Transformers, Hugging Face
Backend: FastAPI, Django, REST APIs
Infra: Docker, Redis, AWS, Azure
CV / NLP: OpenCV, YOLO, PaddleOCR, spaCy

• MPhil in Artificial Intelligence, PUCIT (2024-2026)
• BS Computer Science, Lahore Garrison University (2017-2021)

• Building production-ready LLM and voice AI systems
• Advancing research in instruction fine-tuning
• Preparing for high-performance engineering roles (AI/ML systems)