Profile banner
Mohit Goyal

Mohit Goyal

ML engineer building agentic AI, RAG, and applied LLM systems

rockerleo-goyal
Delhi, India
Joined July 2026
Favicon imagemohitx.in
Favicon imagegithub.com
Favicon imageleetcode.com
Favicon imagepeerlist.io

Network

2.5K connections

Summary

Builds agentic AI systems and RAG infrastructure that move research ideas into production. Mohit's portfolio and project writeups emphasize engineering around agents (loops, tools, MCP integration, persistence, approvals) and production-ready RAG platforms with citation and observability features. mohitx+3
Publishes and contributes to ML research on parameter-efficient fine-tuning, trust modeling, and agent-related architectures. Mohit is a co-author on multiple preprints and conference papers bridging applied research and implementation. arxiv+2
Active open-source maintainer and public builder: Mohit publishes code and documentation (AgentForge, MemexLLM, model implementations) and writes technical blog posts that detail architecture and engineering trade-offs for production AI systems. github+2
Competitive builder and community participant: public profiles highlight frequent hackathon participation and multiple wins, and Mohit shares lessons from shipping under tight deadlines and from building evaluation and observability tooling for LLM systems. github+1

Work

Education

Projects

Writing

AoT-BTS: An Explainable Age-of-Trust Based Behavioral Scoring Framework for Real-Time Zero Trust Enforcement using SDN

May 1, 2026

Research article presenting AoT-BTS, a temporal Age-of-Trust decay model and explainable trust-scoring pipeline for Zero-Trust enforcement with SDN, including reproducible Mininet evaluations and SHAP-based explainability.

Favicon imageresearchgate.net

GRIT -- Geometry-Aware PEFT with K-FAC Preconditioning, Fisher-Guided Reprojection, and Dynamic Rank Adaptation

January 1, 2026

ArXiv preprint proposing GRIT, a curvature-aware parameter-efficient fine-tuning method that preconditions gradients with K-FAC, reprojects low-rank bases onto Fisher eigendirections, and adapts rank to concentrate capacity where signal resides.

Favicon imagearxiv.org

A Comprehensive Review of Microsoft Semantic Kernel: Architecture, Capabilities, and Research Directions

January 1, 2025

Conference publication reviewing Microsoft Semantic Kernel's architecture, agent-oriented patterns, and research directions (IEEE DELCON 2025).

Favicon imageresearchgate.net

Selected portfolio blog posts

2025-2026

Long-form blog posts on agent engineering and agentic systems (examples include posts on AgentForge development and agentic commerce).

Favicon imagemohitx.in

Hobbies

Active hackathon participant and builder (multiple wins). github+1