I'm Nithin Reddy Poola — an MS Computer Science candidate at UMBC and a systems-focused full-stack & AI engineer. I build the kind of software that has to stay up: platforms that provision real infrastructure, AI products that respond in real time, and developer tools that make engineers faster.
I care about the parts most people skip — failure modes, auth boundaries, observability, idempotency, and the 50ms that separate "fine" from "fast." I treat UX and reliability as the same problem viewed from two ends.
focus: [ AI coding agents, platform engineering, infra automation ]
strengths: [ distributed systems, secure auth (SAML/OIDC/RBAC), real-time AI ]
philosophy: "design for failure · automate the boring · instrument everything"
based_in: "Baltimore, MD · UMBC"
now: "shipping ApeironCode — an autonomous, review-ready coding agent"Real systems, measurable outcomes. Numbers reflect my own projects; values marked (est.) are self-measured estimates.
| Impact | What it took |
|---|---|
| ~90% faster provisioning — lab/VM spin-up cut from ~30 min to under 5 min (est.) | VMware vCenter API automation + templated, self-service workflows |
| 1,000+ concurrent requests handled with ~99.9% uptime (est.) | Idempotent job queues, retries/backoff, health checks, structured logs & metrics |
| 95%+ fewer access tickets (est.) | SAML 2.0 SSO + RBAC, scoped sessions, and audit logging |
| <150ms audio round-trip on a real-time AI interview platform | Streaming pipeline, backpressure handling, low-latency transport |
| ~45% faster LSTM training across a 4× NVIDIA A100 cluster | Custom CUDA kernel optimization + PyTorch profiling |
|
AI coding agent & developer tool An open-source, local-first coding agent that plans, edits, and verifies changes across real codebases while keeping every action inspectable.
|
Full-stack cyber lab platform Self-service portal that provisions isolated lab environments on demand, with production guardrails end to end.
|
|
AI life copilot / multi-agent automation Multi-agent orchestration with guardrails, approvals, replay, and a workflow builder — observability baked into every run.
|
Outbreak forecasting research Forecasting service fusing public health feeds with LSTM ensembles, served through clean analytical dashboards.
|
|
Low-latency conversational AI Streaming audio with live feedback, scoring, transcripts, and analytics — accessibility-first (ARIA) UI.
|
Applied research & infra Federated learning experiments and sensor-data pipelines, with CUDA-level performance tuning on GPU clusters.
|
| Frontend | |
| Backend | |
| AI / ML | |
| Cloud / DevOps | |
| Databases | |
| Systems / Infra | |
| Security / Auth | SAML 2.0 · OAuth2 / OIDC · RBAC · secure sessions · audit logging · secure SDLC |
- ApeironCode — pushing an AI coding agent toward reliable, multi-file, review-ready autonomy.
- Full-stack AI systems — real-time, observable, and production-hardened by default.
- Infra automation — turning manual ops into self-service platforms with guardrails.
I'm always open to working on AI tools, infrastructure platforms, full-stack products, and open source. If you're building something hard, I want to hear about it.
Designed for a 30-second skim and a 3-minute deep read. · © Nithin Reddy Poola