Aman Nambisan

Software Engineer II, Atlan

Mar 2026 — Present · Bangalore

• Migrated an internal agent platform from closed frontier models to open-weight ones, cutting token spend substantially with no measurable drop in task success on internal evals.
• Designed a deep agent that catches anomalous cloud-cost spikes early and works out their root cause.
• Built an AI agent for long-running infrastructure migrations: it chooses deployment rings, tracks progress and works across customer maintenance windows in different time zones with little human intervention, halving the engineering time a migration costs.
• Architected release-management tooling that tracks per-tenant deployment state across the fleet, eliminating a recurring class of top-severity incidents caused by unverified upstream rollouts and unclear ring status.
• Scaled Argo CD to more than a thousand applications across EKS, GKE and AKS, and cut sync time by about 20 minutes per application through tuning and an Argo CD upgrade.

Software Engineer I, Atlan

Jun 2025 — Feb 2026 · Bangalore

• Owned development and review for Atlan's internal developer portal (Vue.js and Go), used across engineering, sales and customer teams, and established it as the source of truth for tenant metadata.
• Took on-demand infrastructure provisioning from roughly a 50% success rate to above 90% every month for more than six months, by fixing Terraform modules and making bootstrap jobs idempotent; moving the flow to Crossplane took it to 100%.
• Extended the AI support chatbot to resolve most weekly support tickets on its own, generating PrivateLink setups and looking up tenant details.

Associate Software Engineer I, Atlan

Nov 2024 — May 2025 · Mumbai

• Integrated an OpenTelemetry agent shipping the developer portal's logs into ClickHouse.
• Wrote a Temporal workflow that monitors resource quota across clouds with Prometheus metrics and Grafana alerting, catching every provisioning and scale-up failure before quota ran out.

Frontend Intern, simPPL

May 2023 — Aug 2023 · Remote, USA

• Built D3.js visualisations over million-row misinformation datasets drawn from X, Facebook and Telegram.
• Built a Discord-triggered content pipeline that scraped articles, summarised them with an LLM and delivered them as a newsletter to more than 300 subscribers.

GPU-Aware Autoscaler for LLM Inference

A Kubernetes autoscaler that scales vLLM inference pods on KV cache pressure and request queue depth rather than CPU, because on a GPU server CPU is not the bottleneck and rises only after users are already waiting.

Traffic spikes fill the cache, the autoscaler reads that as a leading indicator, KEDA adds a replica and the cluster autoscaler provisions an L4 — so time-to-first-token stays flat through the spike.

Kubernetes, vLLM, KEDA, Terraform, Prometheus · Code for GPU-Aware Autoscaler for LLM Inference↗

Autonomous ML Deployment System

An autonomous research loop inspired by Karpathy's Autoresearch. Drop in a binary-classification CSV, declare its schema, and a closed loop takes over: Claude proposes a change, the system trains it on Kubernetes, and only candidates that beat the current champion get promoted. Accuracy went from roughly 70% to 80% across 20 iterations.

DVC versions the data, Kubeflow Pipelines runs training, MLflow holds the champion/challenger registry, and a FastAPI service is rolled forward via GitOps whenever a new champion is promoted. No human in the loop.

Kubernetes, Kubeflow, MLflow, DVC, FastAPI, GitOps · Code for Autonomous ML Deployment System↗ · Demo video for Autonomous ML Deployment System↗

Go Modular Monolith for Ride Dispatch

A Go modular monolith split by domain — booking, matching, pricing — communicating over Kafka, with asynchronous booking processing and real-time ride tracking.

Redis provides a distributed lock so a rider cannot be double-booked, and Google's S2 library handles spatial indexing for proximity search in the driver-matching service.

Go, Kafka, Redis, S2 · Code for Go Modular Monolith for Ride Dispatch↗