$ whoami --verbose
Sanchith Boddavaram Anand
├─ now MS CS @ University at Buffalo · graduating Dec 2026
├─ doing LiDAR SLAM + HD mapping for a Lincoln MKZ autonomous testbed
├─ also SWE Intern @ Floe Inc · CI/CD + a C++ hardware abstraction layer
├─ before 2 yrs SWE @ Onnet Systems, Bengaluru · cloud storage, live IPTV, CUDA video
├─ teaching CSE 4/589 Networking · CSE 4/531 Algorithms · CSE 250 Data Structures
└─ into things that move bytes fast under real constraintsMost CDNs make edges poll the origin. This one inverts it — the origin watches the filesystem with
fsnotify and pushes HLS segments the instant they land, while a load balancer routes each viewer
on live hardware telemetry and geographic proximity.
flowchart LR
subgraph O["ORIGIN"]
direction TB
FS["fsnotify<br/>directory watcher"]
PP["concurrent<br/>push pipeline"]
FS --> PP
end
subgraph E["EDGE TIER"]
direction TB
E1["edge-01<br/>LRU cache"]
E2["edge-02<br/>LRU cache"]
E3["edge-03<br/>LRU cache"]
end
subgraph L["LOAD BALANCER"]
direction TB
REG["lock-free<br/>node registry"]
GEO["geographic<br/>PoP filter"]
HW["hardware health<br/>scoring"]
REG --> GEO --> HW
end
PP -- "HLS segments" --> E1
PP -- "HLS segments" --> E2
PP -- "HLS segments" --> E3
E1 -. "telemetry" .-> REG
E2 -. "telemetry" .-> REG
E3 -. "telemetry" .-> REG
HW == "routed request" ==> V(["viewer"])
V -- "TLS · HMAC-SHA256" --> E2
Load test, measured with k6 — including the part where it breaks:
| Concurrent users | Throughput | Avg latency | p95 | Error rate |
|---|---|---|---|---|
| 500 VUs | 1,691 req/s | 116 ms | 262 ms | 0.00% ✅ |
| 1,000 VUs | 1,785 req/s | 280 ms | 652 ms | 0.00% ✅ |
| 5,000 VUs | 793 req/s | 725 ms | 2,170 ms | 31.70% |
Saturation lands around 1.8k req/s — and it isn't the application. It's OS socket exhaustion. Knowing exactly where your system falls over is worth more than pretending it doesn't.
Also does mid-stream failover via dynamic playlist rewriting, HMAC-SHA256 request auth, native TLS termination, and an embedded React dashboard streaming live edge telemetry.
🔐 libsodium-pynacl — end-to-end encryption, from research at IISc Bangalore
Client/server E2E encryption built on libsodium via PyNaCl. Came out of a research internship at the Indian Institute of Science, where I implemented a hybrid public key encryption SDK for smart-city sensor infrastructure — cutting per-message transmission overhead 15% against the incumbent scheme.
The production side of that work lives in iudx-encrypted-resource-server
and iudx-python-sdk (India Urban Data Exchange).
⚡ Gitty — switch Git/GitHub accounts on Windows, instantly
PowerShell tool that swaps your Git identity, SSH key, and credential-manager entry in one command. Built it because I kept committing to the wrong account, which is the only good reason to build anything.
🚩 Red-Flag — a Chrome extension that reads privacy policies so you don't have to
Pairs Cheerio scraping with a chunked LLM inference pipeline to surface the clauses you'd regret agreeing to, inline, in real time. Collaborative project — repo lives here.
🚗 Autonomous vehicle research — SLAM benchmarking + HD mapping @ UB
Benchmarked six LiDAR-inertial SLAM stacks — LIO-SAM, FAST-LIO, GLIM, LeGO-LOAM, LOAM — on a novel environment, scoring localization error and drift to pick the production mapping pipeline for our Lincoln MKZ testbed.
Then authored the HD map itself in Lanelet2 and Vector Map Builder — lane geometry, regulatory elements, MGRS-anchored point clouds — which is what got the vehicle driving closed-loop in Autoware.Universe for the first time.
Forks I live in: nebula (ROS 2 LiDAR/radar drivers) ·
LIO-SAM ·
MS-Mapping ·
pc_utm_to_mgrs_converter
timeline
2019 — 2023 : B.E. Information Science, Atria Institute of Technology, Bengaluru
2022 : Research Intern at IISc Bangalore — hybrid public key encryption for smart-city sensors
2023 — 2025 : Software Engineer at Onnet Systems — led a cloud storage platform to 30,000+ users in month one
: Built live IPTV geofencing and JWT auth, plus a CUDA-X video watermarking pipeline
2025 : Moved to Buffalo for an MS in Computer Science — TA for Networking, Algorithms, Data Structures
2026 : Graduate Research Assistant — LiDAR SLAM and Autoware HD mapping on an AV testbed
: SWE Intern at Floe Inc — CI/CD for embedded releases and a C++ hardware abstraction layer
Languages
Backend, infrastructure & observability
Frontend & data
Systems & research
plus CUDA-X · Autoware · Lanelet2 · MinIO/S3 · FFmpeg/HLS · k6 · JWT & HMAC
The snake eats my contributions
