I build deterministic, production-grade multi-agent architectures and enterprise-level evaluation systems. Bridging high-scale async backends (FastAPI) with complex LLM state orchestration to eliminate stochastic drift and cascading tool failures.📍 Addis Ababa (EAT / UTC+3) — Available for global remote engineering roles.
30+
10K+
5yr+
I'm a software engineer specializing in autonomous LLM workflows and full-stack development. My technical focus revolves around bridging advanced intelligence models with high-throughput API architectures and beautiful responsive user interfaces.
Whether implementing custom Model Context Protocol (MCP) integrations, designing secure database routing layers, or executing precise visual experiences, I engineer for system longevity, strict type-safety, and real-world outcomes.
"The difference between static code and dynamic utility lies in architectural intent. I build interfaces that think."
What: Architecting multi-agent frameworks, prompt workflows, and custom MCP integrations.
Why: Transforms static models into active, task-solving business tools with guarded execution limits.
What: Linking secure Python/FastAPI and Node.js backends to performant Next.js & Flutter frontends.
Why: Provides compile-time type safety across the network layer, eliminating standard runtime crashes.
What: Refining backend request lifecycles and frontend Core Web Vitals (LCP, INP, CLS).
Why: Slashes computing overhead while delivering sub-100ms UI responsiveness for immediate user conversion.
Production web & mobile platforms
Active digital platform consumers
Post-migration type-safety standard
Agent automation pipelines benefit
Cognitive Stack & Tech Arsenal
Advanced autonomous AI orchestrations alongside a production-ready, type-safe full-stack ecosystem spanning frontend platforms, cloud infrastructure, databases, and automated pipelines.
32+ Technologies Mastered
I do not wait for a backlog or granular specifications. I operate directly at the intersection of systems architecture, customer alignment, and product vision—autonomously delivering production-grade backend systems and full-stack integrations.
I drop straight into complex, distributed codebases, reconstruct accurate system models via AST parsing, and start resolving business bottlenecks immediately without disrupting your core team.
Uncertain systems break under real-world traffic. I enforce strict Pydantic/TypeScript schemas at the API boundary, guard AI outputs with comprehensive evaluation graphs, and treat latency as a core metric.
I own the entire delivery loop—from writing high-throughput NestJS microservices and configuring RabbitMQ channels to compiling custom React Native/Flutter layouts and database indexing.
Selected Work
Real-world projects spanning AI copilots, mobile ride-hailing, platform infrastructure, and high-impact experiments.
FastAPI sales automation engine for synthetic-prospect outreach, combining inbound email/SMS webhooks, HubSpot write-back, Cal.com booking, enrichment signals, Langfuse traces, and tau2-bench evaluation.
“Built the sales agent with sink-routed outbound safety and sealed-eval guardrails before any live-channel workflow.”
Sales-domain benchmark and trained critic layer for Tenacious-style B2B outreach, with contamination-aware task generation, deterministic scoring checks, preference data, and a small LoRA judge.
“Turned prospect-facing sales failures into a measurable benchmark and judge gate instead of trusting generated drafts by default.”
Schema integrity and lineage attribution system that turns inter-system dependencies into formal contracts, detects schema/type/statistical drift, and reports downstream blast radius.
“Made AI pipeline contracts executable: validate, attribute, and explain exactly which consumers break when data drifts.”
PDF triage and extraction pipeline that detects document origin, layout, and domain, escalates extraction strategies by confidence, builds PageIndex trees, and answers with provenance chains.
“Converted messy PDFs into auditable structured facts with confidence gates, retrieval indexes, and source-cited answers.”
Event-sourced lending pipeline for document intake, extraction, credit analysis, fraud, compliance, and decision orchestration over an append-only ledger.
“Modeled lending decisions as replayable event streams so every agent action has an audit trail.”
Fork and evaluation workspace for DAB, a realistic enterprise data-agent benchmark spanning multi-database integration, messy joins, unstructured text transformation, and domain knowledge.
“Used a serious external benchmark to stress data agents against multi-database enterprise complexity rather than SQL-only toy tasks.”
View all 17 projects and case studies
Deep-dives analyzing core statistical boundaries, machine learning objectives, and mathematical formulations in production AI systems.
A mathematical first-principles exploration of paired vs. unpaired bootstrap sampling distributions for LLM evaluation. Maps why pairing is correct by experimental design in within-subject setups, explains the variance-reduction mechanism, and details empirical simulation results in Python.
Direct Preference Optimization (DPO) bypassed complex reward models, but SimPO introduces a length-normalized, reference-free objective that mitigates length bias. This deep dive maps their objective functions, gradient deviations, and VRAM footprints under LoRA configurations.
Client Testimonials
"Natnael transformed our customer support tooling in record time. The experience was seamless, collaborative, and the end result exceeded expectations."
Kate Rogers
Product Designer @ ABC Corp
Project
Support tooling rebuild
Impact
faster internal handoffs
30+
projects shipped
10K+
users reached
65%
bug reduction
Career Path
Building the future of AI and mobile through years of dedicated engineering and product vision.