4+ years shipping · 11 production systems · 2 companies started · 0 problems sent back unsolved
I'm not a task-taker. Give me a problem — I'll come back with a shipped solution.
At Levo.ai I joined as a founding engineer before the product was defined. I talked to enterprise customers, found the gaps, and built the infrastructure they now depend on daily: kernel-level traffic sensors, API security engines, agentless discovery tooling, and a CLI built from zero. At IndiaMART, I ran ML search infrastructure at millions-of-queries-per-day scale.
I've also tried to build 2 companies. Both failed. Both taught me more than any job ever has.
Here's what 4+ years of owning problems end-to-end looks like:
| What | Outcome | |
|---|---|---|
| 🔬 | eBPF sensor with OpenSSL + BoringSSL TLS inspection (Levo.ai) | TLS visibility across 60% of customer environments |
| 🪟 | Windows PCAP sensor as a persistent Windows service (Levo.ai) | +30% threat detection coverage across enterprise clients |
| 🔌 | Native IIS HTTP/HTTPS kernel filter for enterprise IIS (Levo.ai) | +40% observability — wire-level traffic without touching app code |
| 🛡️ | GraphQL security engine — auth, batching, introspection, injection (Levo.ai) | REST + GraphQL + SOAP under one security engine |
| 🛠️ | API security CLI, built 0→1 (Levo.ai) | Terminal-first: auth flows, traffic uploads, scan triggers |
| 📄 | Agentless HAR-based API discovery (Levo.ai) | +20% customer adoption by removing the agent install requirement |
| ⚡ | Flask → FastAPI migration across services (Levo.ai) | 45% latency drop, 2× request throughput |
| 📡 | Signoz observability rollout across all applications (Levo.ai) | 30% faster incident resolution across the team |
| 🗄️ | Redis cache layer for ML search (IndiaMART) | 40ms → 15ms — 62.5% improvement at millions-of-queries scale |
| 🤖 | ML model + search API infra (IndiaMART) | 98.9% uptime, +20% search accuracy, queries 30% faster |
| 📦 | SQL data pipeline automation (IndiaMART) | 90% of extraction automated, 60% manual effort eliminated |
Most engineers write code. Fewer try to build products. I've done both — and failed — which is worth more than you'd think.
Build Verge Inc — AI-Powered AEC Platform (with Stanford engineers)
Built an NLP + BIM platform to automate construction workflows using LLMs — generating rooms, walls, doors, voids programmatically from language prompts. Ran into product-market fit and funding walls. The research lives on.
Washbe — On-Demand Campus Laundry
Launched a React Native-based laundry platform with real-time scheduling. COVID hit. The market changed faster than we could pivot.
Both failed. I learned what you can't learn from any job: how to define a problem without a PM, talk to customers without a script, make architectural bets with no safety net, and keep shipping even when it's not working.
That's the founder mindset. That's what I bring to an engineering team.
Languages
APIs & Backend
Infrastructure & Cloud
Data & Storage
Observability
Low-Level & Specialty
Most production work lives in private Levo.ai repos. Public activity is open tooling, side projects, and experiments.
type Focus struct {
Topic string
Why string
}
currentWork := []Focus{
{"eBPF + XDP", "Wire-level packet processing — no kernel modules, no overhead assumptions"},
{"Agentic Debug Tooling", "AI-assisted root cause analysis for live production incidents"},
{"API Behavioral Drift", "Detecting when APIs deviate from their established production baseline"},
{"OpenTelemetry Internals", "What tracing frameworks actually do — and what they still miss"},
}I'm looking for a team that's building something hard — where the problems aren't solved yet, the architecture isn't locked in, and engineers are expected to think, not just execute.
If that's you, let's talk.
