Senior engineer with 14+ years of experience across enterprise software, cloud infrastructure, and AI integration — including work with PwC and KPMG. I specialize in backend systems that are observable, scalable, and production-ready, from secure LLM guardrails to sub-$10 ETL pipelines processing millions of records.
profile = {
"name": "Asad Tungeker",
"location": "Chicago, IL",
"experience": "14+ years in enterprise IT",
"focus": ["LLM / GenAI Integration", "RAG Architectures", "Cloud-Native APIs", "Data Engineering"],
"stack": ["Python", "C# / .NET", "TypeScript", "Azure", "FastAPI", "LangChain"],
"open_to": "Senior / Staff Engineer roles in AI, Backend, or Data Engineering — USA",
}prompt-shield-AI · Python LLM Guardrails
Dual-layer guardrail system protecting LLM apps against prompt injection, jailbreaks, and PII leakage before inference hits the model.
auto-refund-ai · Python Streamlit RAG
NLP-powered complaint processor that classifies sentiment, extracts intent, and triggers automated refund decisions.
langGraph-superstore-AI-assistant · LangGraph Python
Stateful multi-turn shopping assistant with deterministic agent workflows for inventory, orders, and returns.
async-document-processor · C# .NET Azure Functions OCR
Async pipeline converting 10,000+ legacy scanned contracts to structured Word/Excel via OCR + OpenAI — 95% accuracy.
adSpend-predictor · Python scikit-learn
Sales predictor using polynomial regression with feature transformation and overfitting diagnostics, exposed via Streamlit.
job-title-canonicalization · BERT PyTorch Contrastive Learning
Fine-tuned BERT with triplet loss to map 10,000+ raw job titles to canonical seniority levels using LLM-generated hard negatives.
entity-resolution-dedupe-pipeline · Sentence-Transformers Active Learning
Customer deduplication pipeline using cosine similarity + active learning — 92% precision at scale.
| Domain | Achievement |
|---|---|
| ETL at Scale | ADF + .NET pipeline normalizing 3.3M records to Parquet + Cosmos DB — run cost ~$10 |
| AI Observability | Full per-call LLM traceability (prompts, responses, datasets, versions) for regulated audit compliance |
| Model Compression | BERT → DistilBERT distillation — 10× parameter reduction, 97% accuracy retained |
| Legacy Migration | Monolith → DDD microservices with TDD; BrainCloud → Cosmos DB for real-time analytics |
| LLM Safety | Production guardrail blocking injection, jailbreaks, and PII at the inference boundary |
- Certified Generative AI Specialist (CGAI)
- Microsoft Certified Data Engineer
- Sitecore XM Cloud Certified Developer
Open to Senior / Staff Engineer roles in AI, Backend, or Data Engineering across the USA
