Open to opportunities

Sushrut Barmate

@sushrutbarmate

Message

AI/ML engineer building fast RAG pipelines and production-ready LLM backends.

India

Message

What I'm looking for

I want to build AI systems end to end—from data pipelines and RAG retrieval to reliable, test-driven LLM backends that run safely for concurrent users.

I built a lightweight data ingestion pipeline at LeMiCi IQ that extracts structured distributor metrics from getdistributors.com and processes 70,000+ records into clean CSVs without heavy browser automation.

On the same team, I engineered single-agent and multi-agent RAG pipelines for contextual retrieval, pushing end-to-end responses to under 2 seconds and improving retrieval accuracy by 24% with dynamic chunking and prompt orchestration.

At IIT Jodhpur, I worked on predictive modeling for EEG-TMS brain signal forecasting using Echo State Networks, including preprocessing and time-series pipelines for 50+ GB of data with NumPy and PyTorch.

I also built ForgePilot, a constrained autonomous code generation system with deterministic validation, pytest-driven self-healing, and a stateless multi-tenant FastAPI backend that sandboxes concurrent runs in cloud deployments.

Experience

Work history, roles, and key accomplishments

AI/ML Engineering Intern

LeMiCi IQ

Sep 2025 - Dec 2025 (3 months)

Designed a lightweight data ingestion pipeline to extract structured distributor metrics, reducing data collection latency by 40%. Engineered single-agent and multi-agent RAG pipelines for knowledge retrieval, achieving sub-2-second response times.

Data Ingestion RAG Prompt Engineering

Undergraduate Research Assistant

IIT Jodhpur

Aug 2024 - Jan 2025 (5 months)

Developed predictive models using Echo State Networks for EEG-TMS brain signal forecasting, achieving 83% predictive accuracy. Implemented preprocessing and time-series learning pipelines using NumPy and PyTorch.

Echo State Networks NumPy PyTorch Time Series Analysis EEG