Backend Engineer • AI Agent Engineer • Systems Builder
I build production‑grade backend systems and AI agents that actually scale. Real users, real traffic, real edge cases.
I work at the intersection of backend engineering, data systems, and applied AI. My job is turning messy, real‑world requirements into clean, reliable, cost‑efficient systems — the kind that survive contact with production.
Lately that means multi‑agent LLM systems: tool‑calling agents, hybrid retrieval pipelines, streaming chat, and the unglamorous plumbing (token budgets, memory, evals) that makes them dependable instead of impressive‑once.
I care deeply about:
- ⚡ Latency & performance
- 🧱 Reliability & maintainability
- 💸 Cost‑awareness (tokens, API credits, compute)
- 🧠 LLM realism — context limits, failure modes, hallucination control
I’m engineering‑first, not hype‑first.
The questions I ask before writing a line of code:
- How do we handle 2MB+ of text without blowing the context window?
- What happens when the schema changes tomorrow — or doesn’t exist yet?
- Can this run 10× bigger without 10× the cost?
- When the model is wrong, how does the system fail — loudly, safely, or silently?
I’d rather ship a boring system that holds than a clever one that breaks at scale.
Backend & Infra — FastAPI · Django · MongoDB (async) · Pinecone · WebSocket streaming · async batch & background workflows
AI / LLM Systems — LlamaIndex (ReAct agents, multi‑agent handoff, tool dispatch) · multi‑provider orchestration (Claude, GPT) · hybrid retrieval (keyword + vector + reranking) · token‑aware chunking & budgeting · agent evaluation & optimization loops
Automation & Scraping — Playwright (JS‑heavy & infinite‑scroll) · stealth, pagination & rate‑limit handling · external data‑API integration
Real repositories that represent how I work
🔧 Backend & Infrastructure
- 3dvt-backend — Python backend powering a 3D visualization platform
- django-minio-backend — Django + MinIO backend for scalable object storage
🌐 Frontend‑Connected Systems
- 3dvt-frontend — Frontend paired with a production backend
- web-selector-annotator — Browser extension for DOM selection & annotation
🧪 Tools & Utilities
- office-word-count — Replacement utility for deprecated office tools
- tekber3101 — Practical local development & experimentation project
Recurring themes across my work:
- Handling large, messy, unstructured datasets
- Designing agent tools for unknown or evolving schemas
- Making AI systems predictable, observable, and cost‑aware
I’m a strong fit if you need someone who can:
- Design AI agents that don’t hallucinate themselves to death
- Bridge product needs ↔ backend constraints ↔ LLM limits
- Own systems end‑to‑end — architecture, retrieval, evals, and the parts nobody likes to maintain
I work especially well with:
- Startups moving fast but needing stability
- Teams serious about AI beyond the PoC stage
- Anyone wrestling with high‑volume or unstructured data