Projects
Some are polished. Some are still duct-taped. All are mine.
🛠️ End-to-End Logistic Regression MLOps Pipeline
Role: ML Engineer (Personal Project)
Duration: Started July 2025 — still ongoing
URL: GitHub – mlops
Built from scratch (yes, with pain and glory) — a full MLOps regression pipeline focused on the “real-world mess”:
- CI/CD automation
- Experiment tracking with MLflow
- Data & model versioning via DVC
- Drift detection
- Cloud-native deployment with Kubernetes
Currently adding more chaos (Airflow, monitoring, maybe a robot arm). Trying to simulate production without actually breaking production.
📚 Curriculum LLM Platform
Role: Core Developer (Collaborated with Project Owner)
Duration: Feb 2025 – Jul 2025
Also built from scratch — a lightweight content delivery platform with:
- Payload CMS (Node.js), Next.js frontend
- Python FastAPI backend
- Multi-env Docker setup (local/dev/staging) with GitHub Actions
- Modular RAG pipeline (because static content is boring)
Designed for edge inference on Intel XPU devices — think “LLM-on-the-go” kind of vibe.
⚠️ Internal tool, so no public repo. But yes, it does work and didn’t explode.
🧠 Enterprise GenAI Platform (Intel OPEA)
Role: Core Developer (under Tech Lead Direction)
Duration: Nov 2024 – Jan 2025
URL: GitHub – GenAIExamples
Optimized containerized GenAI inferencing on Intel iGPU/dGPU (XPU):
- Debugged docker container hell
- Migrated workflows from
systemd
toDocker Compose
- Reduced setup friction across teams
- Built internal PoC demos to show off inferencing
🔍 AI Workload Validation (Intel WSF)
Role: Validation Support Engineer
Duration: Jul 2024 – Oct 2025
URL: GitHub – Workload Services Framework
Worked on validation support for AI workloads on Intel Xeon platforms:
- Wrote and ran test cases
- Helped triage bugs using JIRA (my best frenemy)
- Helped stabilize and cover more edge cases in platform testing
Learned that testing is 30% logic, 70% caffeine.
🎨 Final Year Project – Colorizing Grayscale Images with Deep Learning
Role: Researcher
Duration: Oct 2022 – Jul 2023
More info: LinkedIn Project
Built a conditional GAN model (“Colour Stack cGAN”) to colorize grayscale images:
- Literature review + model architecture from scratch
- Trained using paired datasets (gray ↔ color)
- Evaluated perceptual quality and accuracy
- Use cases: restoring old photos, automatic media colorization, digital nostalgia
Presented at academic symposiums, and no one fell asleep — a win.
✳ This page will self-update whenever I break/ship something cool.