Open to opportunities

Ayush Jangir

@ayushjangir

Message

I build production agentic AI systems that turn complex planning into auditable decisions.

India

Message

What I'm looking for

I'm looking to build production AI systems where agent autonomy, deterministic reasoning, evaluation, and human oversight create measurable business outcomes.

At Honasa Consumer Limited (Mamaearth), I build and ship production agentic AI systems that replace hours of manual planning with governed, auditable workflows.

I own an eight-step multi-agent media-planning pipeline using Google ADK, LangGraph, Gemini, BigQuery, and human-in-the-loop checkpoints. I design deterministic tool-use boundaries so every numeric output is traceable to computation rather than generated by a model.

I also build RAG pipelines over FAISS vector stores, Bayesian Media Mix Models, and causal measurement experiments, including 20+ GeoLift experiments across 70+ Indian cities. Earlier, I built Power BI dashboards and analytics models at Honasa before converting from an internship to a full-time role.

My research work with LNMIIT and Cleveland Clinic included a PyTorch inference pipeline for cardiac CT landmark detection that achieved 92% accuracy across 500+ scans and reduced latency by 40%. I hold an M.Tech in Computer Science with a specialization in AI and ML.

Experience

Work history, roles, and key accomplishments

Current

AI/ML Engineer

Current

Honasa Consumer Limited

Jun 2026 - Present (3 months)

Design and ship autonomous, self-orchestrating AI agents in production, including an 8-step multi-agent pipeline. Built RAG pipelines over FAISS vector stores and owns the full AI agent lifecycle from discovery to deployment.

Google ADK LangGraph Gemini RAG Faiss BigQuery Python SQL Vertex AI Bayesian Marketing Mix Modelling GeoLift Causal Inference Human In The Loop Versioned Evaluation Sets

Intern, Analytics

Honasa Consumer Limited

Dec 2025 - May 2026 (5 months)

Built Power BI dashboards and ran media mix and geo-lift models on syndicated retail data. Converted to a full-time role after six months based on performance.

Power BI Media Mix Modeling GeoLift Data Analysis

Computer Vision Research Intern

LNMIIT × Cleveland Clinic

Jun 2025 - Dec 2025 (6 months)

Designed a phase-independent PyTorch inference pipeline for a 3D deep learning model for cardiac CT landmark detection, achieving 92% accuracy and cutting inference latency by 40%.

PyTorch Deep Learning Computer Vision 3D Medical Imaging Inference Optimization