$ cat about.md
# SAAD SHARIF AHMED — AI ENGINEER / PM

I build LLM systems
that survive production. 

> AI Transformation Lead @ Think Studio
> Founder @ Aiqarus LLC — agent platform for regulated industries
> Previously: Antler EIR · Agoda · weDevs

> Domains: agentic systems, RAG, LLM training (SFT/QLoRA/SimPO),
>          enterprise compliance (SOC2, HIPAA, EU AI Act).
> 10+ yrs. 4 countries. 45+ shipped.
./view_work.sh → mail saad@
$ ./bench --metrics

Key metrics

40%
SLA reduction
AI workflow automation at Agoda
verified ✓
85%
AI adoption
Tool adoption within 60 days
verified ✓
~6×
Cost reduction
Smart model routing at Aiqarus
verified ✓
45+
Projects delivered
Across SaaS, travel, agencies
verified ✓
$ ls -lah ./work

Featured work

2025 [Product]
2025 [Product]
6-Product SaaS Launch
Led product strategy and GTM for 6 WordPress SaaS products at weDevs. Achieved 6-figure MRR with 6000%+ ROI on paid channels. Built
Product Strategy GTM WordPress SaaS
2025 [AI & Agents]
2025 [AI & Agents]
Multi-Agent Research System
Collaborative AI research system using CrewAI with specialized agents for market analysis, competitive intelligence, and synthesis. Features shared memory and task delegation.
CrewAI LangChain Perplexity Python
2025 [AI & Agents]
2025 [AI & Agents]
Auto-Deploying Agent Platform
Next-gen agent platform with proactive task execution, persistent memory, evaluator-as-judge logic, and containerized deployments. Full execution traceability for enterprise environments.
LangGraph OpenAI Docker Python
2025 [AI & Agents]
2025 [AI & Agents]
Enterprise RAG Knowledge System
Production-grade RAG architecture with hybrid search, semantic chunking, and multi-modal retrieval. Integrated with vector databases for enterprise document intelligence.
LangChain ChromaDB FAISS FastAPI
2025 [Automation]
2025 [Automation]
LLM-Powered Workflow Engine
Intelligent automation system for intent detection, ticket classification, and auto-routing. Built at Agoda to streamline partner operations with AI-driven decision making.
GPT-4 n8n Python REST APIs
2025 [Automation]
2025 [Automation]
AI Partner Deck Builder
Autonomous agent that generates customized partner presentation decks using market data, performance metrics, and brand guidelines. Reduced manual deck creation time by
OpenAI Python GCP Slides API
$ kaggle competitions list --mine

Competitions and research

$ tree ./skills/

Skills

├─ ai_engineering_agents
  • LLM APIs (OpenAI, Anthropic)
  • Agentic pipelines, tool-calling
  • RAG, vector DBs
  • LangChain / LangGraph / CrewAI
  • MCP, A2A, Google ADK
├─ llm_training_alignment
  • SFT, QLoRA, SimPO/DPO/KTO
  • On-policy data, rejection sampling
  • LLM-as-judge, eval harnesses
  • GGUF, vLLM serving
  • Qwen3, LLaMA
├─ programming_infra
  • Python, TS, SQL, Bash
  • React, Next.js, Tailwind
  • Docker, Vercel, Modal, AWS, GCP
  • FAISS, Chroma, Weaviate
  • CI/CD pipelines
├─ compliance_governance
  • SOC 2, HIPAA, GDPR
  • EU AI Act, DSA
  • Zero Trust, audit trails
  • Human-in-the-loop systems
  • AI governance frameworks
├─ data_analytics
  • SQL, Spark, Scala, HiveQL
  • Metabase, Tableau, Superset
  • GA4, GTM, A/B testing
  • Funnel & cohort analysis
  • Success metric design
├─ ai_transformation
  • Enterprise adoption programs
  • Change mgmt, workflow audits
  • P&L, pricing, GTM
  • C-suite & board reporting
  • Certified ScrumMaster®
$ cat ./credentials.txt

Professional certifications

[SCRUM ALLIANCE]
Certified Scrum Master (CSM)
id: 001260903
$ ls -t ./writing | head -3

From the blog

[JUN 02 2026]
Postiz + Facebook Pages: fixing “Can’t load URL: domain not in app’s domains”

If you self-host Postiz and try to connect a Facebook Page, you can slam straight into Facebook’s “domain

read →
[MAY 12 2026]
I queried the WordPress.org public APIs for 63,418 plugins and 8,168 themes. Here’s what the data says.

Three days of WordPress.org public-API queries. 27 charts. The hidden monopolies, the PHP gap, and the 2025 AI

read →
[APR 28 2026]
I Trained a 4B Agent on Five Actions. It Only Learned One.

I trained Qwen3.5-4B on five enterprise agent actions: call a tool, answer directly, clarify, escalate, refuse. The model

read →
$ ls ./writing/ # view all posts →
$ ./chat_with_saad --model=ai_saad@v1

Chat with AI Saad

~/chat · ai_saad@v1 · openai · hf-spaces ● online
# ai_saad@v1 ready. ask me about my work, stack, or the time I broke prod with a bad regex.