Latency is the enemy of retention. Complexity is the enemy of scale. Ambiguity is the enemy of production AI.
I design systems where technical architecture directly serves business execution.
My work sits at the intersection of:
- Production GenAI
- Rust-based backend systems
- Fintech and regulated workflows
- Agentic automation
- KYC / KYB / AML / LCB-FT operations
- Founder-grade product engineering
I do not believe in “AI demos” that collapse the moment reality touches them.
I care about:
- schema-first outputs
- deterministic workflows
- audit trails
- human validation
- observability
- evaluations
- cost control
- boring reliability
Because yes, the model can be magical — but the system around it still needs to behave like adults built it.
The first wave of agents was fragile Python scripts wrapped in hope.
The next wave will be:
- typed
- observable
- permissioned
- auditable
- async by default
- integrated with real business systems
- safe enough for regulated environments
That is why I am increasingly building with Rust, Axum, Tokio, Postgres, OpenTelemetry, and agent frameworks such as Rig.
| Area | What I Build | Why It Matters |
|---|---|---|
| Remolab | Venture studio infrastructure for AI, fintech, and deep-tech products. | Helping founders move from idea to serious technical execution without drowning in architecture debt. |
| CaseReady / RavenKYC | Supervised AI assistant for KYC, KYB, AML and LCB-FT teams. | AI prepares. Analysts validate. Audit trails prove. No black-box compliance theatre. |
| AI Reading Club | A public learning space for foundational LLM papers, from transformers to interpretability. | Turning research papers into practical engineering intuition. |
| Agentic Operating Systems | Internal AI workforce layers for research, sales, CRM, follow-up, scheduling, and execution. | Replacing scattered automation with governed, role-based agents. |
| Rust AI Systems | From-scratch ML and transformer implementations in Rust. | Learning the hard parts properly instead of outsourcing understanding to notebooks. |
| Venture / Project | Technical Work | Product / Business Outcome |
|---|---|---|
| Welcome Place CTO / Product Officer |
Led product and engineering for fintech workflows serving migrants and newcomers in Europe, including regulated onboarding, KYC/KYB, operational workflows, and platform architecture. | Built financial-access infrastructure in a sensitive, regulated environment where reliability, trust, and compliance matter. |
| Remolab Co-founder / CTO |
Architected and mentored 0→1 AI, fintech, and deep-tech products across product strategy, system design, infrastructure, and execution. | Helped early-stage teams turn vague ambition into shippable systems. Annoyingly useful, apparently. |
| CaseReady / RavenKYC | Designed a supervised AI workflow for blocked KYC/LCB-FT cases: missing data detection, extraction checks, information request preparation, analyst validation, and auditability. | Targets the expensive middle of compliance operations without pretending AI should make final regulatory decisions. |
| AI Reading Club | Built a GitHub-based curriculum around foundational LLM papers, with practical sessions, notes, and Rust-oriented explorations. | Creates a sustainable learning loop between ML theory, engineering practice, and community discussion. |
| Agentic Sales / Ops Stack | Designed AI agents for lead research, outreach preparation, CRM hygiene, follow-up coordination, and executive support. | Turns scattered manual operations into repeatable, governed workflows. |
- LLM orchestration
- RAG systems
- schema-first outputs
- structured JSON generation
- evaluations and CI gates
- prompt regression testing
- observability and tracing
- human-in-the-loop workflows
- safety boundaries for autonomous agents
- Rust 2024
- Axum
- Tokio
serdethiserrortracing- OpenTelemetry
- Postgres
- async workers
- strongly typed domain models
- newtype-driven architecture
- KYC / KYB
- AML / LCB-FT
- audit trails
- risk workflows
- compliance operations
- analyst validation
- ACPR-aware product thinking
- vendor risk and operational control
- PostgreSQL
- Supabase
- Fly.io
- Docker
- GitHub Actions
- S3-compatible object storage
- OpenTelemetry
- Langfuse
- Qdrant
- Terraform
- AWS / GCP patterns
I run and maintain the AI Reading Club, a GitHub-based reading club around foundational LLM papers.
Current themes include:
- transformer architecture
- attention mechanisms
- BERT interpretability
- decoding and generation
- data foundations
- scaling and efficiency
- fine-tuning and alignment
- implementing ML ideas from scratch in Rust
Repository:
github.com/hghalebi/ai-reading-club
Because apparently reading papers was not painful enough, so I decided to add Rust.
- Rust-native AI agents
- supervised compliance automation
- human-in-the-loop KYC / AML workflows
- local-first AI infrastructure
- agentic CRM and sales intelligence
- AI coding agents and production delivery
- transformer internals from first principles
- evaluation-driven GenAI systems
- private agentic operating systems for founders
I am interested in serious conversations around:
- production GenAI systems
- fintech and regulated AI workflows
- Rust backend architecture
- AI agents for operational leverage
- KYC / AML automation
- venture studio technical execution
- AI product strategy from prototype to production