Building intelligent systems end to end — from swarm algorithms to edge hardware to cloud pipelines.
Previously active at github.com/hadyy17 — consolidated and restructured here.
I'm an AI/ML Research Engineer based in India, working at the intersection of machine learning research, embedded systems, and distributed AI. Research Fellow at IIT Bombay (returning) — multi-robot systems & swarm dynamics. Currently exploring LLMs, edge AI, and IoT security.
I believe in building things that matter — AI that reaches people who need it most.
- Large Language Models — inference, fine-tuning, agents, RAG
- Swarm Intelligence & Multi-Robot Systems — emergent behavior, particle dynamics
- Edge AI & TinyML — model deployment on constrained hardware
- IoT Security — hardware-rooted cryptography, PUF-based authentication
- AI for Healthcare — applying ML to longitudinal medical data
Firmware, simulation & analysis for a swarm of custom differential-drive robots — Active Brownian Particle dynamics, Run-and-Tumble motion, and emergent collective behaviour. Research conducted at IIT Bombay.
Arduino Embedded Systems Physics Simulation Python nRF24L01
Comparative NLP study — TF-IDF baselines, Bi-LSTM sequence models, and DistilBERT fine-tuning for binary misinformation classification. Config-driven, reproducible experiments with full evaluation suite.
NLP BiLSTM Transformers HuggingFace Python
Fine-tuning LLaMA 2 using PEFT/LoRA for domain-specific instruction following. Parameter-efficient training on consumer hardware.
LLaMA2 LoRA PEFT HuggingFace Python
End-to-end Retrieval-Augmented Generation pipeline using Mistral 7B with vector search for grounded, citation-backed responses.
Mistral RAG Vector Database LangChain Python
High-throughput LLM inference using vLLM — PagedAttention, continuous batching, and optimized serving for large language models.
vLLM Inference PagedAttention Python
Hardware-rooted security framework for constrained IoT devices. SRAM Physical Unclonable Functions for device identity, Ring-Oscillator TRNG for session key entropy, AES-128-GCM authenticated encryption. Targeting IEEE IoT Journal 2026.
ESP8266 ESP32 RaspberryPi C++ Python Cryptography
🛡️ GhostKey
Predecessor to HardSec-IoT. Encryption keys born from physical hardware noise — never stored, never transmitted in plaintext.
ESP8266 AES-GCM SRAM-PUF IoT
🧩 NemoDuo
Hybrid multi-agent system — Nemotron 3 Super (120B) as strategic Planner + Nemotron 3 Nano (4B) running locally as Executor. Custom agent loop without LangChain, Streamlit dashboard, Docker Compose setup.
NVIDIA Nemotron Multi-Agent LLM Python Streamlit
Benchmarking suite for Nemotron model family evaluation.
Benchmarking LLM-Evaluation Python
| 🏢 HSBC Technology | Software Engineer — distributed ML, multi-cloud pipelines, 10M+ asset inventory at scale |
| 🎓 IIT Bombay | Research Fellow — 50-unit microbot swarm, emergent behavior, particle motion dynamics |
| 🔬 LD College of Engineering | Research Intern — NLP misinformation detection, IoT security |
| 🤖 DD Robocon 2021 & 2022 | Software Team Lead — ABU Robocon, National |
AI/ML: TensorFlow · PyTorch · HuggingFace · scikit-learn · LangChain · RAG · PEFT · Distributed Training (Horovod)
Embedded & Robotics: ESP8266 · ESP32 · Arduino · Raspberry Pi · Jetson Nano · AES-GCM · SRAM PUF · RO-TRNG
Cloud & Data: GCP · AWS · Azure · BigQuery · Spark · FastAPI · Docker · Kubernetes
Languages: Python · C++ · Bash · R · MicroPython
- 🌍 Open to AI/ML Research Engineer roles internationally
"The best way to predict the future is to build it."