About
I am Mohammad Yaser Hussain, a fresh AI and Machine Learning graduate from Jayaprakash Narayan College of Engineering, affiliated to JNTU, Hyderabad.
I work across the full range of AI engineering, from systems-level thinking and performance optimization to designing intelligent pipelines and applied machine learning. I am not attached to any one approach or tool. I follow the problem and figure out what it actually needs.
Building software that works in the real world matters to me more than building software that looks good in a demo. That means caring about performance, reliability and the experience of whoever ends up using it. I am also a published IEEE researcher with work in machine learning applied to real world market analysis.
Project Experience
- Privacy-first on-device engine running Llama, Phi, and Qwen LLMs alongside Vision-Language Models entirely offline on 2 to 3 GB RAM hardware
- Custom C++17 JNI bridge integrating llama.cpp, Google MediaPipe, and TensorFlow Lite for low-latency multimodal text and vision inference
- Asynchronous model loading and memory management via Kotlin Coroutines, ensuring stable inference and a non-blocking UI under heavy computation
- Clean Architecture, Dagger Hilt, and WorkManager powering background hot-swapping of multi-gigabyte model files
- Fully offline Windows desktop AI assistant with Qwen2.5 LLM, multimodal vision, STT via faster-whisper, and TTS with zero cloud dependencies
- Hybrid ML/DL intent pipeline combining LinearSVC and TF-IDF for near-instant classification with DistilBERT (ONNX) for complex queries, routed by complexity score with adaptive CPU/GPU scheduling
- Concurrent PyQt6 application decoupling DL inference, wake-word detection via openWakeWord, and OS system control across isolated worker threads with structured JSONL telemetry
- Semantic memory via ChromaDB and SQLite with spaCy NLP for entity extraction and multi-turn context, alongside real-time audio waveform and hardware telemetry panels
- Multi-agent orchestration via CrewAI Flow coordinating autonomous agents across real-time data fetching, SEC EDGAR filing analysis, and report generation
- Real-time Form 4 and 8-K ingestion via ATOM feeds with Pydantic for type-safe config and BeautifulSoup for robust XML parsing
- Data validation guardrails enforcing integrity, with Google Gemini 2.0 Flash auto-verifying and correcting generated insight reports against raw filing data
- Multimodal deep learning model predicting e-commerce product prices from text and product images, achieving 26.51% validation SMAPE across 75,000 products
- Frozen MiniLM-L6-v2 text encoder (384 dims) fused with frozen MobileNetV2 image encoder (1280 dims) via a 3-layer MLP, keeping trainable parameters at 459K on a 4GB GPU
- Diagnosed and resolved severe underfitting through log1p price scaling and post-hoc linear distribution alignment, lifting predictions from a narrow $4 to $30 band to mean $27.34
- Full inference suite covering an interactive CLI, a Tkinter GUI with threaded model loading, and batch generation scoring 75,000 products in 5.5 minutes
Technical Skills
Languages
AI / ML
LLM & Agentic AI
Android & Mobile
NLP, Vision & Tools
GitHub Statistics
Publications
Peer-Reviewed · Journal Article · 2024
Enlightening Paths: Python's Vision into the Electric Vehicle Market
View on ResearchGate ↗Education & Certifications
Degree
Bachelor of Technology
Jayaprakash Narayan College of Engineering
Affiliated to JNTU, Hyderabad