Applied AI · AI/ML Engineer · Researcher
Tejas Belakavadi Kemparaju
I build ML systems & LLM-powered agents, from production inference pipelines to agentic architectures. M.S. CS from NJIT. Founder of Cnotes.in (50K+ users).
I build ML systems & LLM-powered agents, from production inference pipelines to agentic architectures. M.S. CS from NJIT. Founder of Cnotes.in (50K+ users).
Stateful AI agent for career workflows — resume analysis, job matching, skill gap detection. RAG + FAISS + vLLM.
Synthetic instruction-tuning pipeline for a 15B code LLM. LoRA/QLoRA via PEFT — 70% GPU memory reduction.
Solo-built full-stack CRM for Local Cable Operators. LAMP + React, idempotent billing, role-based auth. 50K users.
Extends TAPE's batch graph construction with an anchor-scored sequential merge pipeline. Novel edge-retention policy and ILP feasibility formulation for constrained LLM agent planning under tool-guided adaptive execution.
Ensemble ML applied at particle physics scale. CatBoost + LightGBM with advanced feature engineering — 99% accuracy on CERN electron mass collision prediction.
Fully convolutional neural network for early-stage retinopathy screening. 93% accuracy, end-to-end clinical usability with TensorFlow and Django.
Currently being finalized. Will be listed upon submission.
ML pipelines for event data processing (XGBoost, scikit-learn). Real-time inference APIs in FastAPI, sub-200ms latency. Precision/recall evaluation with A/B-style comparisons.
ML pipelines for event data processing (XGBoost, scikit-learn). Real-time inference APIs in FastAPI, sub-200ms latency. Precision/recall evaluation with A/B-style comparisons.
Java backend services on AWS. 40% API latency reduction. Docker + Kubernetes for service reliability at scale.
Solo-built a full-stack CRM (LAMP + React) for Local Cable Operators — scaled to 50K+ users. Idempotent billing, audit-ready event logs, role-based auth.
Looking for AI Engineer, ML Engineer, Forward Deployment, and Backend roles — especially at AI-native companies.
Say hello →