AI/ML Engineer & Researcher — building reliable spatial intelligence for embodied agents and production ML systems.
-
🎓 B.E. Information Science & Engineering — focused on AI/ML and intelligent systems
-
🔭 I’m currently working on learning-based spatial intelligence and reliable memory for embodied agents
-
🌱 Deepening skills in machine learning, computer vision, graph neural networks, and robot learning
-
📚 Research experience in graph neural networks and fraud detection, with an IEEE Xplore publication
-
💬 Ask me about spatial memory, scene graphs, graph neural networks, or computer vision
-
📫 Reach me at prashanths272005@gmail.com
-
⚡ Fun fact I take my coffee seriously, everything else lightly.
- 🧠 reliable-spatial-memory — Embodied agent with a confidence-weighted spatial scene graph, multi-sample verification policy, and CLIP-based language grounding. Evaluated over 7,720 runs across 5 policies, 4 scenarios, 2 noise models, and a 4-policy language ablation. Python, NumPy, NetworkX, OpenCV, CLIP, spaCy.
- 🚀 production-ml-service — Production-shaped ML service with FAISS semantic search, Redis caching, and a backend-agnostic LLM client (mock / Groq / self-hosted). Ships with a retrieval evaluation harness across three architectures (dense, hybrid, hybrid+rerank: NDCG@5 = 0.922), Prometheus alert rules, structured logging, a 5-container Docker Compose stack, and a full-stack CI smoke test. Python, FastAPI, FAISS, Redis, MongoDB, Docker, Prometheus.
- 📄 Adaptive Heterogeneous Temporal Graph Neural Framework for Fraud Detection in DeFi — IEEE Xplore (Scopus-indexed).
