Charef Bahria
Senior Full Stack & AI Engineer
Senior full-stack engineer, 5 years in, specialising in production LLM systems and SaaS products I own end to end. I built an agentic contract-review pipeline that ran in production for over a year (about 70% of contracts needed no human edit) and a Stripe-style API-key platform that a product with 100k+ monthly requests adopted over its own. TypeScript, NestJS, React/Next.js, event-driven GCP.
Based in Berlin (Chancenkarte, Blue Card eligible). Can start part-time immediately, full-time once the Blue Card is issued.
I'm a senior full-stack and AI engineer who owns products end to end, from an ambiguous brief through architecture to production. Lately that means production LLM systems: I was the primary engineer on ThatAPICompany's first AI project, a contract-review pipeline that went from a one-month feasibility proof to over a year in production.
Most of my work is TypeScript: NestJS services on event-driven GCP, React/Next.js frontends, and SDKs other developers integrate against. At ThatAPICompany that was a lean agency team with no dedicated ops or QA org, so testing, monitoring and handoff were part of the job, not someone else's.
Open source: in LangChainJS I added .doc support to DocxLoader, with tests and docs, merged (PR #7421).
Based in Berlin (Chancenkarte, Blue Card eligible). Can start part-time immediately, full-time once the Blue Card is issued.
Languages
Senior Full Stack Software Engineer (Remote)
Software agency and SaaS studio. Led end-to-end delivery of multiple products and client engagements, owning scope, architecture and delivery in a lean team.
Full Stack Software Engineer (Remote)
E-commerce platform; cross-functional team of 12.
BSc Computer Science, University of BoumerdesSep 2018 - Jun 2021
Thesis: ARIMA demand-forecasting models in Python on real sales data for inventory planning.
Agentic contract review for a legal-tech client
Reviews third-party contracts against the client's legal playbook. A Gemini-powered agentic flow (LangChain, RAG) generates redlines, reviews its own output and retries; anything still uncertain is flagged for a lawyer (human in the loop). LangSmith tracing and evals gate each release.
Stripe-style API keys and SDK, as its own product
Multi-tenant auth and API-key management SaaS, built 0-to-1 and owned end to end: NestJS backend, React dashboard, rate limiting, TypeScript SDKs modelled on Stripe's, and a BigQuery usage-analytics pipeline.
Gemini-powered photo analysis pipeline
Two decoupled NestJS services running an async pipeline on GCP. Gemini analyses dog photos (breed, body condition score, other analytics) and generates AI portraits. Pub/Sub-driven, with automatic retries and a dead-letter queue so nothing is silently lost. Results are enriched with breed and health data from TheDogAPI/TheCatAPI, with fallbacks so an upstream failure doesn't degrade the result.
Salesforce integration for a medical imaging parts client
REST API connecting distributors to Salesforce. DDD with CQRS cleanly separates inventory/pricing queries from order creation and tracking. JWT auth via TheAuthAPI. Versioned across multiple customer deployments. GCP Cloud Run with GitHub Actions CI/CD.
Marketing sites with interactive API demos
Multilingual (EN/ES) marketing sites in SvelteKit with interactive API demos, pricing and signup flow. i18n with slug-based translation keys for easy content updates. Static generation to Vercel, adapted for both cat and dog API branding.
Open source contribution
Added .doc support to DocxLoader, with tests and docs, merged (PR #7421).
I use Claude Code daily. My loop: I write the plan and constraints, the agent implements, I verify against the code, tests and the running app. The value is in the review: on a recent take-home the model proposed cosine similarity as a “fit score” — that measures wording, not eligibility, so I removed it — and it left a log line that would have written personal data into cloud logs, so I removed that too. Faster model for implementation, stronger one for planning and second opinions.
I worked closely with Charef for four years at That API Company and watched him grow from a strong individual contributor into one of our most trusted senior engineers. Over that time, he took on full technical ownership of complex work — on our legal-contract redlining AI project, he owned the scope, stakeholder management, and delivery, moving from executing defined tasks to making the architecture and trade-off decisions the rest of the team depended on. On client work, he operated end-to-end: he led our Salesforce integration from scoping through build, communicating directly with the client throughout. He also raised the bar on code quality across the team and was a genuine mentor — his code reviews made the engineers around him, myself included, better. His seniority is well earned: he handles ambiguity independently and consistently lifts the people around him. I'd gladly work with him again.
I write about systems-level engineering: the kind of implementation details that most tutorials skip over. These are deep dives into topics I found genuinely confusing, written so they don't have to be.
A deep dive into TCP fundamentals and building an HTTP server from scratch, covering the socket layer most engineers never touch.
A production guide to replacing a managed API gateway with a self-hosted Nginx setup: real tradeoffs, config details, and lessons learned.
I am open to senior full stack and AI engineering roles in Berlin and remotely. If you are building something technically interesting and want someone who ships end to end, get in touch.
charafbahria@gmail.com