A sub-5MB ‘Detect-Split-Classify’ Computer Vision pipeline for automated rice grain quality detection. Built with SAM2 for mask generation and INT8 QAT for on-device edge deployment.
AI/ML Engineer from research to deployment.
I design and build production AI/ML systems — custom Computer Vision pipelines, NLP engines, and agentic workflows — as an Independent Researcher & Consultant. 5+ years shipping with Python, PyTorch, and the full-stack web.
A Lead AI/ML Engineer & independent researcher.
I'm a Lead AI/ML Engineer and Independent AI Researcher & Consultant. I design and build end-to-end machine learning systems — bridging cutting-edge research with production-ready deployment across computer vision, NLP, and agentic AI.
My flagship independent projects include Hybrid99 — a sub-5MB Edge AI Computer Vision pipeline for automated grain quality detection — and Resume TLM, a three-phase NLP information extraction engine fine-tuned on DistilRoBERTa-Base. Before going independent, I led AI engineering at Kuanta, where our agents reduced startup-evaluation time by 70%.
I care about precision, clean architecture, and code that scales. Currently obsessed with edge-deployed inference, INT8 quantization, and high-accuracy document parsing.
Where I've shipped.
ML Engineer
Independent AI Researcher & Consultant·AI Research · Architecture · ConsultingLed AI architecture for Innova OS (Innova Amtech), built two flagship ML pipelines — Hybrid99 (Edge AI CV) and Resume TLM (NLP) — through independent research, and engineered backend systems as a consultant for Thinkuvate.
Lead AI Engineer
Kuanta AI·Venture IntelligenceEngineered AI research agents that automated team assessment and multi-stage market validation, cutting manual analysis time by 70%.
Sr. Full Stack Developer
Weblianz Solutions·Software SolutionsArchitected enterprise MERN applications and integrated chatbots / RAG pipelines for diverse client verticals. Led migration to NestJS.
Founder
Scaptor·Product & ConsultingBootstrapped a full-service web agency. Delivered 20+ projects on Next.js, Flutter, GCP, AWS — managed the full discovery-to-deploy lifecycle.
Things I've built.
A curated slice of recent work in AI, full stack, and product.
A modular PyTorch NLP pipeline for high-precision resume parsing. DistilRoBERTa + GLU spatial fusion + CRF head. 66M params, ~78ms CPU — no LLM at inference.
Modular agentic AI engine for startup pitch-deck analysis, orchestrating Team / Finance / Market scoring agents on Celery + Redis.
AI-powered outreach platform — generate personalized emails and research 6,000+ companies with retrieval-grounded context.
Education-management platform for schools, colleges, and libraries. Multi-tenant, edge-deployed on Cloudflare + Supabase.
Type-safe ORM for Cloudflare D1 — fluent query builder, relationships, migrations, and lifecycle hooks built for the edge.
CV Score AI
AI-powered CV scoring system using Gemini + GPT-4o + vector DBs. Real-time feedback with 90% match accuracy on JD pairing.
AI Knowledgebase
Custom RAG pipeline from scratch — chat with PDFs and scraped sites. Qdrant for retrieval, Gemini for reasoning, context-aware memory.
EmergenCare Platform
Healthcare marketplace connecting providers and patients. Real-time medical records and high-concurrency scheduling on NestJS.
Biometric Web Identity
Full WebAuthn (FIDO2) implementation for passwordless biometric login with hardware-backed session security.
What I work with.
Tools sharpened over five years building and scaling high-performance systems.
AI / ML Research01
Cloud & Data02
Core Dev03
Self-directed, formally trained.
JS, Python & Cyber Security
Computer Hardware & Networking
Senior Secondary (Computer Science)
Let's build something durable.
Open to senior AI engineering and tech-lead roles. Also happy to chat about retrieval pipelines, agent architectures, or your weekend project.
karan.963@icloud.comDownload CV →