README.md

metadata

title: README
emoji: 🦅
colorFrom: indigo
colorTo: blue
sdk: static
pinned: false

HawkEyes AI

Production AI for retail-execution, identity, and conversational systems

---

About

HawkEyes Digital Monitoring Ltd. builds production computer-vision and generative-AI systems for enterprise retail-execution monitoring, mobile-financial-services branding audits, biometric identity verification, and real-time conversational AI. Models and Spaces published here power live services for clients including BAT, Unilever, and Nagad, covering 2M+ retail outlets at ~99% detection accuracy.

Flagship repositories

Repo	Description
BAT_Master	Central production CV+analytics service for a BAT retail-execution program — campaign-specific YOLO detection/segmentation, planogram compliance, face recognition.
UBL-Backup	Unilever Bangladesh retail-execution API — 10+ versioned YOLO models across display, POSM, share-of-shelf, and sachet audits.
Voice-AI-Agent	Production real-time conversational voice agent — WebRTC streaming, VAD→STT→LangGraph→TTS, MCP tool-calling, per-session memory.
VITS-FineTune	Custom Bangladeshi-accent Bangla TTS fine-tuning (Coqui VITS + grapheme frontend), built for production call agents.
HE-Sherlock	Strictly-grounded Bangla document RAG — OCR + FAISS + Gemini.
Malay-DAIA	Retail item-detection + agent face-verification API for a Malaysian deployment.
Squere_Demo	Modular multi-audit retail API — Display Audit, POSM, Sachet, Share-of-Shelf.
OCR_CloudVision_Gemeni	Hybrid OCR pipeline — Google Cloud Vision + Gemini + NLTK normalization.
distance-calculate	Detection-driven object distance/spacing estimation for retail placement analytics.

What we work on

• Retail-execution computer vision — POSM detection, planogram/shelf compliance, share-of-shelf, competitor detection.
• Biometric identity — face verification for agent/KYC-style checks.
• Conversational AI — real-time voice agents and production RAG/agentic systems.
• Low-resource Bangla speech & OCR — Bangla TTS and document intelligence.
• MLOps — GPU-optimized async inference, containerized deployment, self-hosted on-prem GPU infrastructure.

Contact

Maintained by MD Rakibul Hasan Naym, Head of AI — naym.mj@gmail.com · Portfolio