AI Engineer | Software Engineer | LLM & MLOps Researcher
Research Assistant @ Micro Electronics Research Lab (MERL)
LFX’25 Mentee @ RISC-V International
Open-source contributor focused on production-ready AI systems, LLM evaluation, and reproducible ML pipelines.
I’m an AI Engineer and Researcher working at the intersection of LLMs, MLOps, and open-source systems.
I build reliable, testable, and deployment-ready AI pipelines rather than experimental-only models.
- 🧠 LLM Evaluation & Benchmarking — functional, syntactic, adversarial
- 🛡️ Hallucination Mitigation — GAN-based approaches for private LLMs
- ⚙️ Reproducible ML Pipelines — CI/CD, logging, SLA-aware validation
- 📊 RISC-V Tooling & Data — machine-readable specifications and verification
💡 Making AI systems trustworthy in production is my passion.
-
🔹 Research Assistant — MERL
LLM evaluation pipelines, benchmarking frameworks, RISC-V tooling -
🔹 LFX’25 Mentee — RISC-V International
Machine-readable RISC-V specifications, schemas, and CI validation
Languages: Python · Scala · Verilog · Java · Shell · JavaScript · HTML · CSS
AI / ML: PyTorch · TensorFlow · Hugging Face Transformers · GANs · LLM Evaluation · NumPy · Pandas · Scikit-learn
MLOps & Engineering: CI/CD · Docker · REST/gRPC · Logging & Monitoring · Reproducible Pipelines · Git · GitHub Actions · Linux · pytest
Data & Config: JSON · YAML · MySQL
🛡️ AI4org — GAN-based Hallucination Mitigation for Private LLMs
- Built a privacy-first ML pipeline to detect and mitigate hallucinations in private LLMs
- Designed a GAN-style generator/discriminator for hallucination detection
- End-to-end pipeline: ingestion → validation → reproducible training → containerized inference
- Integrated CI/CD, automated testing, and monitoring for production readiness
📌 Designed for enterprise and on-prem LLM deployments where reliability matters.
🔬 ArcheV — LLM Benchmark Suite
- Engineered a reproducible LLM benchmarking framework
- Standardized JSON I/O and CI-driven evaluation pipelines
- Validates functional and syntactic correctness for deployment decisions
📘 RISC-V Unified Database
- Maintained versioned YAML/JSON schemas for RISC-V tooling
- Implemented CI validation to ensure data integrity and observability
- Improved downstream reliability for tooling and ML pipelines
- 🎓 Linux Foundation Mentorship Program (LFX) 2025
- 🧪 Research Assistant at MERL
- 📊 Improved LLM benchmarking reliability by ~25%
- 🧠 Hands-on experience with LLMs, GANs, MLOps, and CI/CD
- 📝 Contributor to open-source and research-grade tooling
- 💼 LinkedIn: Shehroz Kashif
- 📧 Email: sharooz57@gmail.com
⭐ If you find my work useful, feel free to star a repository.
🤝 Open to collaborations in AI, LLMs, MLOps, and open-source systems.