DEEPKERNELS STUDIOS
HOLA,
I'M
KARANDEEP
SINGH
Skill: Agentic Architecture & Orchestration
I design and deploy multi-agent pipelines where LLMs do not just respond, they plan, delegate, and execute. Using LangGraph and CrewAI, I build systems with specialized agents across research, ops, and data that collaborate autonomously with human-in-the-loop checkpoints where it matters. No toy demos. Production-grade.
SYSTEMS
Skill: Contextual Retrieval Engineering
I build RAG pipelines that actually retrieve the right chunks using semantic chunking, contextual embeddings from Voyage AI, and hybrid vector search across MongoDB Atlas and Qdrant. From enterprise document Q and A to real-time research agents, the goal is accuracy you can measure and latency you can benchmark.
OPTIMIZATION
Skill: LoRA / QLoRA, PEFT, and Domain Adaptation
Trained on real healthcare data across skin-lesion detection and GAN-based synthesis, I fine-tune open-source models to outperform base systems on niche tasks. The work is grounded in domain adaptation, lower inference cost, and reduced hallucination rates for specialized use cases, with IEEE-published research behind the healthcare claims.
FROMIDEATOREALITY-Design and deploy multi-agent pipelines with tool calling, memory, and human oversight.
-From research agents to autonomous ops workflows, full architecture through production.
-Integrations across CRM, databases, APIs, Slack, and email.
-Contextual RAG pipelines with semantic chunking, hybrid vector search, and LLM-powered Q and A.
-Vector databases across Qdrant, MongoDB Atlas, and Supabase pgvector.
-Built for enterprise docs, research bases, and real-time knowledge retrieval.
-LoRA and QLoRA fine-tuning on domain-specific datasets.
-Specialty in healthcare, medical imaging, and tabular-data-adjacent workflows.
-IEEE-backed research on GAN-based augmentation and domain adaptation.
-Next.js plus FastAPI plus LangGraph across frontend, backend, and the AI layer.
-Production deployment with Cloud Run, Redis caching, and Supabase SQL.
-From MVP to scalable backend with semantic caching and real users.
I'm Karandeep Singh, an AI Engineer focused on production-grade agentic systems, applied AI research, and full-stack AI products. My work spans IEEE-published healthcare research, national hackathon wins, and building DeepKernels into a platform used by 10K+ people.
(About me)
