Hi, I’m Annas.
Building intelligent systems. Engineering them for production.
AI Engineer at BIK GmbH
I design and build scalable backend systems, distributed architectures and AI-powered software, from system design to production deployment.
- Backend Engineering
- Distributed Systems
- AI Infrastructure
I build LLM platforms that make it past the demo.
Tenant isolation, encryption, grounded retrieval and evaluation, from the first prototype to production.
Multi-agent systems
Orchestrated agents for retrieval, research and automation, with A2A and MCP.
Knowledge graphs and RAG
Neo4j graphs, semantic chunking and hybrid retrieval that ground every answer.
Secure LLM platforms
Per-client isolation, database-level encryption, migrations and Docker deployments.
Selected work
AI systems
Products and systems I built, from live translation to agents over millions of rows.
90% accuracy over 7M+ rows
LangGraph, FastAPI, Arize PhoenixReal-time voice assistant
92% transcription accuracyMore experiments
- AI Chef AssistantMeal planning, recipes, grocery automation
- Arabic NLP: sentiment and NER9+ models compared, GPT-4 against AraBERT
- Network intrusion detectionBinary and multi-class, NSL-KDD
- Algorithmic bias mitigationAIF360, reweighing, fairness metrics
- Medical image classificationCNN and ResNet transfer learning
- Automated MLOps pipelineVersioning, drift detection, retraining
Four roles in production AI
-
Nuremberg
AI Engineer, BIK GmbH
- Architected multi-tenant isolation for an enterprise AI knowledge platform on Azure: 10+ data domains split into per-client databases, with versioned schema migrations across every tenant.
- Designed database-level envelope encryption at rest, with key-rotation tooling and a phased production rollout.
- Built multi-agent orchestration for retrieval and question answering, grounded in a Neo4j knowledge graph with entity resolution and graph-aware semantic chunking.
- Owned Docker deployments, per-tenant LLM cost tracking and GDPR deletion across every store; 30+ merged PRs.
-
Islamabad
AI/ML Developer, Stixor
- Led the design of agentic AI for contract automation and self-editing agents.
- Built a RAG legal chatbot with model versioning, tracing and evaluation.
- Built FastAPI backends for Malakah, a legal AI startup that raised $600K in seed funding.
- Cut inference latency and cost with hybrid retrieval, caching and monitoring.
-
Remote
AI Engineer, Niblon
- Designed and deployed 10+ multi-agent and RAG systems with adaptive memory.
- Built LLM-as-Judge and human-in-the-loop evaluators that raised accuracy by 25%.
- Made retrieval 40% faster with quantized embeddings, adaptive chunking and hybrid ranking.
- Researched agent-to-agent (A2A) and MCP-based reasoning for automation at scale.
-
Remote
Associate ML Developer, Developers Den LLC
- Built hybrid RAG (dense FAISS plus sparse BM25) that improved answer relevance by 35%.
- Shipped voice agents with Whisper and ElevenLabs: 10K+ conversations a month at 92% transcription accuracy.
- Connected 50+ SaaS tools to LLM agents, cutting manual work by 60% and embedding costs by 45%.
From demo to dependable
At BIK GmbH I build the foundations of an enterprise AI knowledge platform: isolation, encryption, knowledge graphs and cost tracking for each customer.
Before that, three years shipping agentic and retrieval systems for legal tech, automation and voice. I’m also completing a Master’s in Artificial Intelligence at BTU Cottbus-Senftenberg.
M.Sc. Artificial Intelligence
BTU Cottbus-Senftenberg
B.Sc. Computer Science
HITEC University
Generative AI with LLMs
DeepLearning.AI and AWS; LangChain for LLM Apps
English, Urdu, German
C1, native, A2 and improving