Building Intelligent Systems·Exploring Mamba & SSMs·Engineering Agentic AI·Generative Audio & CV
I build AI systems by going beneath the API — understanding the architecture, implementing the system, experimenting with it, and measuring what actually works.
I'm an Artificial Intelligence & Data Science undergraduate focused on building intelligent systems across Generative AI, State-Space Models, Agentic Workflows, and Computer Vision.
My technical focus centers on:
Mamba State-Space Models Flow Matching Text-to-Speech Agentic RAG Knowledge Graphs Computer Vision
I enjoy working on problems where the value lies not simply in calling an endpoint, but in understanding how the model processes representations, how the orchestration engine is structured, and how to make the pipeline computationally efficient.
⚡ "If it can be automated with a state machine, it should be."
|
Full-Resolution State-Space Text-to-Speech A research-oriented generative audio architecture built around Bidirectional Mamba-2 + Flow Matching.
Stack: |
Multi-Source Autonomous Research System A full-stack research assistant designed around intelligent query routing, GraphRAG, and self-correcting retrieval. User Query
│
▼
LangGraph Agent (Query Router)
├── Semantic Vector Search (BGE-M3 / Qdrant)
└── Knowledge Graph Traversal (Entities/Hops)
│
▼
Reranking & Context Assembly
│
▼
Structured Generation (LLM)
Stack: |
|
Single-Class Anchor-Free Object Detector from Scratch An end-to-end ground-up implementation and architectural study of the Fully Convolutional One-Stage (FCOS) detection framework.
Stack: |
|
Generative AI
│
├── Text-to-Speech (TTS)
├── Continuous Flow Matching
└── Neural Audio Synthesis
State-Space Models
│
├── Mamba & Mamba-2 Architectures
├── Selective State SSMs
└── Linear-Complexity Sequential Modeling
Agentic AI
│
├── LangGraph State Orchestration
├── GraphRAG & Entity Triplet Extraction
└── Multi-Source Autonomous Retrieval
Computer Vision
│
├── Anchor-Free Object Detection (FCOS)
├── Real-Time Edge Monitoring (MediaPipe)
└── Geometric Deep Learning & Quality Inspection
| Domain | Technologies & Frameworks |
|---|---|
| AI / Deep Learning | PyTorch PyTorch Lightning TensorFlow Hugging Face Transformers Mamba / Mamba-2 CNNs |
| Generative AI & Agents | LangGraph LangChain GraphRAG Flow Matching BGE-M3 Ollama ONNX |
| Computer Vision & Data | OpenCV MediaPipe FCOS NumPy Pandas Scikit-learn SciPy |
| Backend & Infrastructure | FastAPI PostgreSQL Qdrant Supabase Docker Git Linux |
| Languages | Python C++ SQL Bash |
Nov 2025 — Dec 2025
- Evaluated modern deep learning architectures including Transformers and CNN-based detection networks.
- Benchmarked deployment viability and optimized inference pipelines for production readiness.
- Improved target precision to 42% through structured dataset curation and systematic metric tracking.
Jul 2025 — Aug 2025
- Developed computer vision solutions for industrial quality inspection and automotive safety.
- Built a real-time Driver Monitoring System (DMS) using MediaPipe, maintaining ~40 FPS on CPU.
- Analyzed production assembly metrics to identify anomaly patterns across vehicle components.
-
B.Tech in Artificial Intelligence & Data Science
SIMATS Engineering, Chennai · 2023 — 2027
GPA:9.1 / 10· Academic Topper Award ×2 -
Certifications
- NPTEL: Deep Learning for Natural Language Processing
- NPTEL: Deep Learning for Computer Vision
- NPTEL: Python for Data Science —
84%(Elite) - Oracle: Database Management Systems (DBMS) —
79%
┌─────────────────────────────────────────────────────────┐ │ CURRENT HORIZONS │ ├─────────────────────────────────────────────────────────┤ │ │ │ 🧠 State-Space Architectures (Mamba / S4) │ │ 🔊 Generative Audio & Flow Matching Optimization │ │ 🤖 Deterministic State Routing in Agentic AI │ │ 🕸️ Knowledge-Graph-Augmented Retrieval (GraphRAG) │ │ ⚡ Quantized Local Inference (ONNX / Edge Hardware) │ │ │ └─────────────────────────────────────────────────────────┘
