Mandip Adhikari

Mandip Adhikari

I work at the seam between AI/ML systems and the infrastructure that keeps them honest in production — data pipelines, identity, search evaluation, and the math that makes models actually scale.

Open to backend, full-stack & infrastructure rolesHattiesburg, MS
mandip.dev…
colophon →

01Experience

SitusHawk

Jun 2026 — Aug 2026

Software Engineering Intern · Huntsville, AL

SvelteKit · TypeScript · Python · Azure Container Apps · Cosmos DB · YOLOv8

  • Built an end-to-end tax-discovery product — SvelteKit/TypeScript front end, Python pipelines on Azure (Container Apps, Cosmos DB, Blob Storage) — that reconciles multi-sensor vessel evidence against a $372M county assessment roll, surfacing 300+ untaxed vessels.
  • Wrote a constant-memory streaming decoder that screens ~10M daily radio messages (430 GB of raw NOAA AIS feeds) down to per-vessel position histories with a bit-level MMSI fingerprint filter, validated at 96.7% accuracy against reference data.
  • Shipped the computer-vision and ML evidence layer: YOLOv8 detection resolving 1,100+ vessels per satellite scene, presence-pattern classification, and a third-party valuation API — turning assets nobody could see into assessable property.

Aquaview

Feb 2025 — Present

Lead Software Developer · Institute for AI Applications — University of Southern Mississippi

Python · GCP Cloud Run · Terraform · OpenTelemetry · OAuth2 / OIDC · MongoDB · GDAL

  • Lead a 9-engineer team on an 8-service platform serving 300K+ datasets — 30+ production releases in 5 months — while driving the Next.js → Vite migration, OpenTelemetry with 99.9% SLOs, Terraform drift guardrails, and merge-gating CI.
  • Shipped a production OAuth2/OIDC identity provider from scratch — PKCE auth-code flow, RS256-signed ID tokens, SHA-256 hashed token storage with MongoDB TTL expiry — plus a JavaScript SDK for one-import popup / modal login.
  • Built a streaming statistics pipeline (Welford's algorithm + reservoir sampling) over 10M+ observation datasets, with complexity-scored routing between sync requests and async Cloud Run Jobs, and automated depth-profile, time-series, and trajectory previews.
  • Deployed a cloud-optimized raster tile server on Cloud Run for COG / STAC / Zarr data, with tuned GDAL caching (HTTP/2 multiplexing, 200MB curl cache) and 4 custom scientific colormaps.

Illumibot

Oct 2025 — Dec 2025

Software Engineering Intern · ArroyoDev

Python · FFmpeg · Mobile APIs · Inference Services

  • Profiled and optimized the computer-vision segmentation pipeline for projector workloads — a 40% cut in processing overhead, turning a laggy mask-generation engine into something that runs in real time.
  • Built a frame-accurate video transition system on FFmpeg xfade / acrossfade filters with seamless multi-clip chaining.
  • Hardened the mobile ↔ inference API layer with request batching, timeouts, and retry logic — a 30% reduction in service failures and predictable end-to-end latency.

02Selected Work

// how I build

  • Measure before you optimize.
  • Make it idempotent.
  • Make it observable.
  • Make it boring — boring survives.
  • Make it hold at 3am.

04Contact

Open to backend & infrastructure roles and genuinely hard systems problems. The fastest way to reach me is email.

Or send a message