I bridge the gap between Data Science innovation and Production reliability.
10+ years of engineering experience · 3.5+ years in MLOps · Enterprise Retail AI
| 🏗️ Building | MLOps & Edge AI Governance platforms for enterprise retail |
| 🎓 Fellowship | AI Performance Engineering @ Nebius Academy |
| 🦀 Learning | Rust — for high-performance inference pipelines |
| 💬 Ask me about | Azure ML · Databricks · Model Governance · Observability · Cost Optimization |
| 📫 Reach me | sunilkunchoor@gmail.com |
|
Automated Model Governance as Code A GitHub Actions "Gatekeeper" enforcing quality gates on every PR — validating code quality, security (Snyk/Semgrep), and model performance before anything reaches production.
|
Seamless Service Integrations for Airflow Production-grade custom Airflow plugins for integrating with Azure and cloud services, with automated evaluation hooks and full MLflow lineage tracking.
|
Unified Observability for ML Pipelines Lightweight telemetry bridge instrumenting every stage of a GitHub Actions ML workflow — metrics, traces, and events to Dynatrace without vendor lock-in boilerplate.
|
👁️ Retail-LensAI-Powered Smart Shelf Vision Computer vision system empowering store associates to instantly identify out-of-stock items, misplaced products, and incorrect price tags — reducing shelf compliance issues in real time.
|
CloudNative Hackathon · Nebius Academy Serverless benchmarking suite that automatically spins up isolated model inference tasks in AWS Lambda containers to profile latency, cold-start, and memory drift.
|
|
LLM Lifecycle Management with "Prompts as Code" End-to-end LLM pipeline with automated evaluation loops — GPT-4 acts as a judge to score prompt quality, enabling data-driven prompt engineering at scale.
|
|
AI Workloads at the Edge Benchmarking study of small form factor hardware devices managing AI workloads deployed at the edge — covering inference latency, power, and model accuracy tradeoffs.
|
Automated QA using Classification Fully automated QA analysis engine using Classification & Clustering — replacing manual defect triage with ML-driven categorisation at scale.
|
☁️ Cloud & Infrastructure
⚙️ MLOps & Data Platforms
🤖 LLMOps & AI Frameworks
📊 Observability & Governance
🔧 CI/CD & Languages
MLOps is about Engineering Rigour, not just writing scripts.
I operate on four non-negotiable principles that guide every platform I build:
| Principle | What It Means in Practice |
|---|---|
| 🎯 Zero-Friction for Data Scientists | Abstract away Kubernetes, Docker, and infrastructure entirely. Data Scientists should think in models and mathematics — not YAML files. |
| 🏗️ Structure Begets Speed | Ad-hoc scripts don't scale to 50+ models in production. Enforcing strict project templates and CI/CD contracts makes every deployment repeatable, audit-ready, and automated. |
| 🛡️ Guardrails Enable Confidence | Strict governance (like the Traffic Light system) lets teams deploy faster — not slower — because safety is baked in, not bolted on. |
| 💡 Frugal Architecture | Cloud costs should not grow linearly with model usage. Optimised inference (ONNX, quantization, edge deployment) is not a nice-to-have; it's a first-class engineering concern. |
| Credential | Issuer | Verification |
|---|---|---|
| 🏅 Databricks — Data, ML & GenAI | Databricks | View Badge |
| 🏅 IBM Data Science Professional Certificate | Coursera / IBM | View Badge |
| 🏅 GitHub Actions | GitHub | View Badge |
| 🎓 AI Performance Engineer Fellowship | Nebius Academy | View Badge |
| 🎓 PGP in AI & ML | UT Austin via Great Learning | View Portfolio |
| 🎓 M.Sc. Mathematics | Bangalore University | — |
If you're building serious AI infrastructure and need someone who treats MLOps as a first-class engineering discipline — let's talk.