Arda Açıksöz

AI systems engineer with four years' production experience, from enterprise telecom backends to AI-native products. Builds conversation intelligence and speech AI systems, and the multi-agent platforms that do engineering work around them — designed for production, with graceful degradation, bounded autonomy and audit trails. Works end to end: product, cloud infrastructure, and the privacy and AI-regulation layer underneath.

Full-time. Owns AI product engineering end to end — architecture, implementation, infrastructure and production operations.

• Designed and operate a conversation intelligence platform in production — it joins video meetings, transcribes and separates multiple speakers, and turns unstructured discussion into structured, searchable analysis.
• Work across the speech AI stack — automatic speech recognition, speaker diarization, domain vocabulary priming and semantic search — including resolving a production speaker-attribution failure on overlapping speech that off-the-shelf diarization could not handle alone.
• Build LLM systems that hold up in production: graceful degradation across models when providers fail, sectioned analysis for long-form content, and honest failure states instead of silent ones.
• Designed and run a multi-agent platform where AI workers produce engineering deliverables under explicit governance — isolation between agents and production systems, bounded delegation that cannot loop, per-run cost and audit trails, and a human approval gate before any work is accepted. Paired with a knowledge base that unifies operational state and semantic retrieval, so people and agents work from one source of truth.
• Built an autonomous development pipeline — from approved task to implemented, reviewed and merged code, with human approval on every path to production.
• Led a privacy and AI-regulation audit of the live platform (KVKK, GDPR, EU AI Act) covering cross-border data transfer, retention and verified erasure.

• Co-founded and built a multi-tenant B2B SaaS for field reporting and visit management — web and mobile clients over a shared API, serving organisations with role-based permissions across 17 business domains.
• Architected the platform around enforced contracts: every API surface defined by explicit schemas with permission gating and pagination rules, validated automatically in CI so drift between backend, web and mobile cannot reach production.
• Own the cloud infrastructure end to end — containerised services, managed databases, object storage, identity, secrets management and least-privilege access.
• Ran security remediation after finding live credentials in version control: rotated them, eliminated a shared administrator identity used across CI, QA and the build host, and replaced it with scoped, short-lived federated access.
• Set the engineering standards the whole team works to — human and AI contributors alike.

Enterprise telecom delivery. Joined as Junior Software Developer; promoted to Junior Backend Developer and given team responsibility.

• Delivered backend services on international telecom programmes for Vodafone UK, Vodafone Albania and STC (Saudi Telecom).
• Built distributed backend services and orchestrated long-running business processes with a workflow engine, on a containerised microservice architecture with event streaming.
• Developed custom identity and authentication APIs, including OTP and password-reset flows.
• Led a two-developer backend team — task breakdown, code review and delivery.

• Build AI systems for small and mid-sized businesses — recruitment screening that evaluates and ranks candidates automatically, inbound email triage with generated responses, and document digitisation that turns PDF catalogues into structured, queryable data with automated price analysis.

Software Intern · Mirsis Bilgi Teknolojileri · Mar 2022 – Oct 2022 — contributed to an ERP system; implemented business logic and designed the database schema.

Software Intern · Replik Yazılım · Aug 2021 – Sep 2021 — database integration and user authentication modules for an ERP project.

Speech & Audio AI — automatic speech recognition (ASR / speech-to-text), speaker diarization, keyterm and domain-vocabulary priming, text-to-speech, source separation, audio pipeline engineering.
Deepgram Nova-3, Recall.ai, Fish Audio, Piper, Demucs, ffmpeg.

AI & LLM Engineering — multi-agent orchestration, agent governance and isolation design, prompt engineering, structured output, model fallback strategies, RAG, vector search and embeddings.
Anthropic Claude API, Claude Code & Agent SDK, OpenAI, MCP, pgvector.

Languages — Python, TypeScript, JavaScript, Java, C#, C++, Dart, SQL, Bash.

Backend — FastAPI, Spring Boot, Node.js, Express, Pydantic, Camunda, Keycloak, REST API and contract design, RBAC, microservices.

Frontend & Mobile — React, Next.js, TypeScript, Tailwind CSS, Angular, Flutter.

Data — PostgreSQL, Supabase, MongoDB, Redis, DynamoDB, SQLite, pgvector.

Cloud & DevOps — AWS (App Runner, ECR, RDS, S3, Cognito, Secrets Manager, Lambda, IAM, OIDC), Docker, Kubernetes, Apache Kafka, Cloudflare, Tailscale, systemd, CI/CD.

Security & Compliance — KVKK, GDPR, EU AI Act, threat modelling, secret rotation, incident response, zero-trust network design, data retention and erasure.

BSc Computer Engineering — Bahçeşehir University · Jan 2019 – Oct 2022
Transferred from Mechanical Engineering, Istanbul Technical University · Sep 2017 – Jan 2019

Turkish — Native · English — Advanced (C1, IELTS 7.5)

Anthropic Academy — Claude Code, Claude API, MCP, Agent Skills (in progress)
TÜBİTAK Project Competition · Yale & Harvard MUN · GYLC