Engineer at heart • Exploring AI, robotics & the human experience
An engineering chronicle. Projects exploring AI, systems, and the intersection of technology with human experience.
Production AI with 89% accuracy. Built an 8-metric evaluation framework including BERTScore and RAGAS. Achieved 236% improvement in context precision. $7.8M annual value creation.
[Cohere, LangChain, RAGAS, ChromaDB]
ML inference with Redis caching. Sub-100ms response time, 95% cache hit rate. Production-grade with async processing and horizontal scaling.
[FastAPI, Redis, DistilBERT, Kubernetes]
→ githubAutomated research synthesis. Processes 100+ papers daily from arXiv. Extracts insights and generates personalized digests.
[Python, LangChain, arXiv API, GPT-4]
→ githubResearch on enhancing multimodal reasoning through RLAIF and Direct Preference Optimization. Created first public multimodal CoT preference dataset. Published findings on DPO limitations for vision-language models.
[LLaVA, DPO, RLAIF, PyTorch, HuggingFace]
→ datasetEnd-to-end deployment pipeline. Reduced model serving latency by 70%. Implemented caching, health checks, auto-scaling.
[Kubernetes, Docker, Redis, Prometheus]
Computer vision with 96% accuracy. Multi-class CNN with ResNet50 transfer learning. Real-time inference pipeline.
[TensorFlow, CNN, ResNet50, OpenCV]
Interactive education tool. Step-by-step visualization of transformer attention. Used by 200+ students.
[JavaScript, D3.js, WebGL, React]
→ githubParameter-efficient fine-tuning. Reduced memory by 90% while maintaining 98% performance. Comprehensive benchmarks across models.
[PyTorch, Transformers, PEFT, LoRA]
→ github