Skip to content
View MohdSaleh's full-sized avatar

Block or report MohdSaleh

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
MohdSaleh/README.md
Typing SVG

Ryan Saleh

Building production-grade agentic AI systems that actually work

Portfolio LinkedIn Medium Email


What I do

I design and ship production-grade agentic AI systems — multi-agent orchestration pipelines, Model Context Protocol (MCP) servers, enterprise RAG architectures, and LLM reliability frameworks built to operate at scale.

My focus is on the architecture problems that most teams hit in production and don't know how to solve: agents that fail silently, tool access with no governance, retrieval pipelines that hallucinate at the edges, and orchestration loops with no observable recovery path.

10+ years in software delivery. The last 3 years deep in the agentic layer.


Outcomes I've shipped

System Result
Multi-agent orchestration platform (LangChain + ReAct + Plan-Execute DAG) 50% reduction in agent failure-recovery time
Enterprise RAG pipeline (pgvector + BM25 + Cohere reranking + deduplication) < 2% hallucination rate in production
Constraint-driven allocation agent (LangChain + Genetic Algorithms + OR solvers) 3× recommendation quality improvement
4× production MCP servers (SharePoint, SQL, workflow triggers, RBAC-governed) Zero invalid resource allocations in production
LLM evaluation harness (OpenTelemetry-style tracing + regression checks) Full observability across agent execution paths

What I architect

🔄 Agentic Orchestration Plan-Execute DAGs · ReAct reasoning loops · self-correction mechanisms · Human-in-the-Loop checkpoints · multi-agent conversation patterns · budget guards · deterministic verification layers

🔌 Model Context Protocol (MCP) Production MCP server design · typed JSON Schema contracts · RBAC allowlists · governed tool exposure · secure enterprise system integration

🔍 RAG & Retrieval Systems Hybrid search (dense + sparse) · query routing · contextual reranking · metadata enrichment · PII filtering · source citation grounding · context deduplication

🛡️ LLM Reliability & Evaluation LLM-as-Judge harnesses · adversarial prompt testing · prompt injection defense · structured output validation · circuit breakers · sandboxed tool execution


Stack

Python FastAPI LangChain Docker PostgreSQL MongoDB React Node.js

AI/ML: LangChain · LangGraph · AutoGen · CrewAI · LlamaIndex · MCP · RAG · ReAct · Pinecone · FAISS · pgvector · Cohere · OpenAI · Anthropic Claude


Featured work

Production multi-agent execution protocol featuring a canonical nine-phase orchestration loop, self-validation mechanisms, deterministic budget guards, and goal versioning for sequential multi-goal execution. Drop-in configuration for agentic coding environments.

multi-agent · orchestration · execution-loops · agentic-ai · LLM


Research repository for LLM reliability engineering: adversarial testing frameworks, prompt injection defense mechanisms, and structured output validation schemas. Covers evaluation patterns for high-stakes production deployments.

prompt-engineering · LLM-reliability · adversarial-testing · guardrails · evals


🔌 plugLLM

Unified API abstraction layer for hot-swapping between OpenAI, Anthropic Claude, and local LLM providers. Automatic fallback orchestration and load balancing across providers with a single interface.

LLM · openai · anthropic · provider-abstraction · python-sdk


Experimental framework for autonomous agent workflows implementing persistent memory architectures, structured tool registry patterns, and hierarchical execution loops built on LangChain and AutoGen.

autonomous-agents · memory-architecture · tool-registry · LangChain · AutoGen


Writing

Technical pieces on agentic architecture, LLM reliability, and production orchestration patterns:

All articles →


Open to

Lead AI Engineer · Agentic AI Architect · Principal GenAI Engineer · AI Systems Architect

Companies building serious agent infrastructure — enterprise, MNC, product-led. Based in Kerala, India. Open to remote-first roles globally.

→ View full portfolio & case studies


"The difference between an agent demo and a production agent is everything that happens when the plan fails."

Pinned Loading

  1. loops-engineering-for-coding-agents loops-engineering-for-coding-agents Public

    Python

  2. xcule xcule Public

    JavaScript 7 4

  3. ThinkShift_LLM ThinkShift_LLM Public

    🧠🔀 Seamlessly manage multiple LLM clients to overcome API limitations. tShift_LLM automatically shifts between language models, ensuring uninterrupted AI responses and bypassing rate limits. Perfec…

    Python 1

  4. ai-artifact-hub ai-artifact-hub Public

  5. IRS-prompt-engineering-lab IRS-prompt-engineering-lab Public

    Python

  6. LLM-Finetuning LLM-Finetuning Public

    Forked from ashishpatel26/LLM-Finetuning

    LLM Finetuning with peft

    Jupyter Notebook