Asgard
The gate is stirring...
I'm Vidar. I guard Abdul's work. Come in.
Agents that ship, built on backends that hold.
I'm Abdul Haque, an AI software engineer in Islamabad, Pakistan. I build production LLM microservices and multi-agent RAG systems on Azure. One of them serves 300+ sessions a day for a government healthcare client.
Voyages
Client work, research, and my own product. Synverk has its own page.
- SYN-01 · 2026
Synverk
A taskboard where agents do the work through MCP and only a human can mark it Done.
Next.jsMCPPostgresBunRead the case study → - WRK-02 · 2025–26
AI Therapy Microservice
Event-driven audio pipeline serving 300+ daily sessions for an Australian government healthcare client.
Azure AKSFaster-WhisperPhi-3.5-MiniService BusRead the case study → - WRK-03 · 2025–26
AI Student Companion
Multi-tenant RAG with per-client isolation. Sliding-window memory cut 100K tokens to 10K a session.
LangGraphFastAPIpgvectorAzure DevOpsRead the case study → - WRK-04 · 2025–26
Fraud Detection and Reporting
Real-time pipeline for regulated retail banking that files Suspicious Transaction Reports every 5 minutes.
KafkaStreamSetsPrometheusGrafanaRead the case study → - RSC-05 · 2025
GraphRAG-Causal
Graph-augmented causal reasoning in news. 82.88% F1 on CausalNewsCorpus, on par with fine-tuned BERT-Large.
CrewAI FlowsNeo4jFastAPIarXivRead the case study → - RSC-06 · 2025
IntentLens
Four agents (scraper, intent analyser, search, supervisor) reading 5,000+ customer reviews, judged by an LLM.
CrewAIGroq APIStreamlitRead the case study →
The World Tree
Each branch is a group of tools I use in production.
AI and LLM
Backend
Data and RAG
Cloud and infra
ML, NLP and speech
Messaging and observability
Languages
The Saga
Work and study, newest first.
- Dec 2025 – Present
Associate Software Engineer, AI
DeltaShoppe (Pvt.) Ltd, Islamabad, partner Kermit Tech (Norway)
- Sole backend engineer rotating across three concurrent client engagements in a small AI engineering pod.
- AI Therapy: Faster-Whisper and Phi-3.5-Mini (INT8) on CPU-only 15 GB AKS pods, with a multi-processing worker pool that fixed lock contention and ONNX out-of-memory failures. JWKS token verification, encrypted audio, dead-letter queues, MLflow.
- Student Companion: propositional chunking (about 10:1 compression), tool injector pattern for per-user permission scoping, PII redaction middleware, Azure DevOps CI/CD.
- Fraud Detection: Kafka pipeline with PII masking on 6 fields and 3 Grafana dashboards with CRITICAL alerts.
- Apr – Nov 2024
Data Science Intern
Securely Innovations (Pvt.) Ltd
- Multi-threaded collection pipeline over about 3M automotive parts records, 3x faster than Selenium-based scrapers, by replicating the network-level API.
- Jun – Aug 2024
Computer Vision and Generative AI Intern
DataInsight Lab, FAST-NUCES
- Fine-tuned Stable Diffusion (DreamBooth, LoRA, QLoRA) for interior design generation on 2×T4 GPUs using model parallelism.
- Aug 2021 – Jun 2025
B.Sc. Data Science
FAST-NUCES, Islamabad
- CGPA 3.24/4.0, final-year GPA 3.75/4.0. Dean's List in Fall 2024 and Spring 2025. Final-year project: GraphRAG-Causal.
The Rune Stone
A short note, and the badges I earned along the way.
I'm a backend engineer who ended up in AI because the hard part of an LLM product is rarely the model. It is the queue, the memory, the permissions, and what happens at 3am when a pod runs out of memory.
Right now I'm the sole backend engineer across three client engagements. I like systems that fail loudly, keep humans in charge of the final call, and are written down well enough that the next person can take over.
Generative AI with LLMs
DeepLearning.AI and AWS
Section Leader, Code in Place
Stanford University, 2025. 95% student retention across a global cohort.
Cross the Bifrost
Write to me. The message goes straight to my inbox and I reply by email.