Gersum Asfaw Kifle

AI Agent & Systems Engineer | Lead QA Engineer

Professional Summary

Senior Systems & AI Agent Engineer with 6+ years of professional experience spanning AI agent orchestration, telecom Business Support Systems (BSS), cloud integration, and autonomous quality engineering. Proven expertise in building production-ready multi-agent swarms (LangGraph, MCP), enterprise hybrid RAG platforms, and self-healing test automation frameworks. Master’s degree in Software Engineering with a strong track record of merging deep learning research with production-grade system design.

Technical Skills & Expertise

AI Agents & Orchestration

Multi-Agent Swarm Architectures LangGraph Model Context Protocol (MCP) Playwright Automation Self-Healing Systems Auto-Patching Loops

RAG & Vector Search

Hybrid Retrieval (BM25 + Dense Vectors) Reciprocal Rank Fusion (RRF) Query Expansion (HyDE/Multi-Query) Ollama FAISS Anti-Hallucination Guardrails

Frontend Development

Angular TypeScript JavaScript HTML5 / CSS3

Software Engineering & Cloud

Python (3.12+) .NET FastAPI Node.js PostgreSQL SQLite MongoDB DuckDB Docker Kubernetes Azure Cloud

Testing & Security

API & Microservice Validation Penetration Testing (Burp Suite, Metasploit) AST Code Analysis (Tree-sitter, sqlglot) Playwright PyTest

Professional Experience

10 Academy
Feb 2026 – Present
Addis Ababa, Ethiopia

AI Agent & Evaluation Engineer

  • Multi-Agent Swarm Architecture: Designed and built Project Chimera (Automaton Auditor), a Fan-Out / Fan-In LangGraph swarm replacing single-prompt LLM chains with parallel AST detective nodes and a deterministic Python Chief Justice node enforcing fact supremacy over LLM outputs.
  • Autonomous AI QA Swarm Engine: Architected an end-to-end self-healing QA testing framework using Playwright and MCP tools. Implemented real-time DOM selector fallbacks (CSS → ARIA → Text → XPath) and a secondary LangGraph repair loop that auto-patches broken test specs from stack traces.
  • Multi-DB MCP Routing Engine: Built DataAgentBench (The Oracle Forge), injecting a 3-layer context engine (schemas, domain rules, memory logs) to route natural language queries dynamically across PostgreSQL, MongoDB, SQLite, and DuckDB via MCP.
  • Enterprise RAG Engineering: Built the Doc Refinery Platform (DRP) for Safaricom BSS documentation—featuring heading-aware chunking, hybrid BM25 + dense vector retrieval fused via Reciprocal Rank Fusion (RRF), and post-synthesis lexical grounding to eliminate hallucinations.
  • Cost-Gated Document Pipelines: Developed a 5-stage document intelligence refinery with confidence-gated switches, cascading digital/scanned PDFs from free text parsers to Vision LLMs only when confidence metrics drop below 0.85.
Safaricom Ethiopia
Past Experience
Addis Ababa, Ethiopia

Lead Software Quality Assurance Engineer

  • Led end-to-end QA, performance, and automation strategies across telecom Business Support Systems (BSS), CRM, Case Management, and billing platforms.
  • Architected scalable automated testing suites integrated into CI/CD pipelines, driving high release stability across critical enterprise systems.
  • Oversaw API validation, integration compliance, and cloud-native service reliability.
Information Network Security Administration (INSA)
Past Experience
Addis Ababa, Ethiopia

Exploit Researcher

  • Conducted vulnerability research, penetration testing, and security auditing using tools like Burp Suite and Metasploit to strengthen national software infrastructure.
Awash International Bank
Past Experience
Addis Ababa, Ethiopia

Digital Channels Officer

  • Managed operational support, security compliance, and testing for digital banking channels and core banking services.

Awards & Innovations

1st Place Winner — Cursor AI Hackathon Ethiopia

July 2026

Co-developed TENAx, an Amharic-speaking, voice-first chronic care management platform for patients with diabetes and hypertension featuring spoken AI coaching and 90-day clinical summaries.

MSc. Research — Proactive Crop Disease Prediction

Implemented a Bi-Directional LSTM (Bi-LSTM) model forecasting ginger bacterial wilt prior to infection using NASA microclimate data, achieving 96.08% accuracy and 97.7% precision.

Agricultural Sensory Grid (ASG)

Full-stack MEAN platform integrating real-time regional weather mapping, soil pH machine-learning crop recommendations, and national yield estimation.

Education & Certifications

MSc in Software Engineering

Addis Ababa Science and Technology University

Thesis: Proactive Plant Disease Prediction Using Deep Neural Networks & Microclimate Analytics

BSc in Software Engineering

Addis Ababa Science and Technology University

Microsoft Certified: Azure Cloud & AI Fundamentals