Anwesha KarSystem Software Engineer at NVIDIA | Master’s in CS | Author | SpeakerMaking intelligent transportation possible. Building what’s next.
The short version ↗
UNDER THE HOOD. AHEAD OF THE CURVE.
Before a robotaxi can navigate or a flying car can take off, its hardware and software need to work together. I work on that foundational layer: the systems behind intelligent transportation.
When this layer fails, a device may never start—or stay reliable enough to use. Getting it right means following clues from boot logs to the hardware itself.
My focus: V2X communication, real-time systems and precise, shared time.
A MESSAGE THAT GETS THROUGH.
BIG FUTURES. LOW-LEVEL DETAILS. ↗
Boot. Connect. Stay in sync.
Then build what’s next.
Start with the result.
Open a card for the story, figures and evidence.
Congestion-aware aerial relays help hazard warnings reach more vehicles.
At 250 DENM packets in the simulation.
IEEE VTC-Fall · 2024
Joint scheduling for periodic status updates and urgent hazard alerts.
Reported reduction versus legacy MAC.
Computer Communications · 2024
Better reception when vehicles compete for scarce radio resources.
Paper-reported gain in extreme congestion.
ACM ICDCN · 2023
Why a trusted, shared clock matters when sensors observe the same scene.
A high-level view of timing reliability.
Research study · PTP / gPTP
Three published papers and a timing-reliability research study. Numerical results are simulation findings. pp = percentage points.
Talks, research communities and the people along the way.
Self-Driving Cars & the Smart City Revolution.
A moment from the awards ceremony.
Conversations and connections off stage.
Mentoring hackathon participants at Google HQ in India.
At Google HQ in India for hackathon mentoring.
Industry mentors meet students: sharing perspectives and answering questions.
Connecting with students and the team at Future Forward in Gujarat.
Bringing industry experience into conversations with students.
Returning as an alumna for a CSE Day panel discussion.
Sharing an alumni perspective with the IIT Hyderabad community.
Speaking about physical AI at the STeP-IN Automotive Tech Conference.
A moment on stage, sharing research and ideas.
Research conversations at the International Conference on Distributed Computing and Networking.
The people and conversations beyond the paper.
A moment from the event floor.
Taking the conversation beyond the lab.
Meeting people working across the AI ecosystem.
Ideas travel further through conversation.
Connections that continue beyond the event.
Scroll or swipe to browse · Select a photo for the full view
Five conversations on research, IIT Hyderabad and finding your own path.
29:39Research, campus life and the M.Tech experience, in conversation with RBR.
Watch on YouTube
18:41A second conversation about IIT Hyderabad, on Ravindrababu Ravula’s Hindi channel.
Watch on YouTube
25:14Research, placements and GATE, with Tathagata Dey.
Watch on YouTube
20:45A conversation about the IIT Hyderabad experience on Bangla e Shekho.
Watch on YouTube
26:51Confidence, creativity and experiences beyond academics, with Atharva Tendulkar.
Watch from 11:48Scroll or swipe to browse · Select a preview to watch on YouTube
A few interactive ways to explore the systems behind the story.
Camera, LiDAR and radar observations need a common time reference. My work on gPTP examines why the reliability of that shared clock matters to autonomous driving.
Move the slider to see a simple idea: readings about the same event become harder to align when their clocks disagree.
Concept illustration only. No attack details, implementation data or experimental results are reproduced.All three observations line up on the same timeline.
If every department runs AI, who owns the bill? I’m exploring token budgets as an operating discipline: allocation, attribution, approval and measurable value.
This miniature “Token Trading Floor” makes the trade-offs tangible. Change a workload and a rate profile to see the monthly inference estimate.
Illustrative calculator. Profiles are invented examples, not provider prices or live feeds. Estimates exclude infrastructure, storage, tools, taxes and other charges.USD per million tokens. “In” = input; “out” = output.
$262.50 per department with an equal split.
21% of the illustrative budget used.
I’m interested in the orchestration layer: which agent can act, what it can access, and when a human needs to step in.
OpenClaw and NVIDIA’s NemoClaw ecosystem make that conversation timely. For a sovereign agentic cloud, residency needs to be part of routing and governance.
Illustrative policy routing only. This demo runs no agents, makes no API calls and transfers no data. UAE residency is a design scenario, not a compliance claim.Research-informed ideas I want to explore, build and bring into the room.
Optimising and securing the Precision Time Protocol
I’m exploring how to make Precision Time Protocol (PTP) synchronisation more efficient, reliable and secure. The goal: a shared clock that connected systems can depend on.
Reducing timing error and synchronisation overhead, while strengthening resilience to disruptions and attempts to manipulate timing. Performance and security need to be considered together.
Helping aerial and ground vehicles work together
As flying cars become part of the mobility conversation, I’m interested in how vehicles in the air and on the road can communicate and coordinate smoothly.
Building on my aerial–terrestrial research: timely warning exchange, dependable links and coordinated communication as vehicles move between ground and air. The aim is a more connected view of what’s happening around them.
06 / THE HUMAN IN THE LOOPI work on intelligent transportation systems (ITS), vehicle-to-everything (V2X) communication and real-time PTP synchronisation.
My focus is the infrastructure that lets connected vehicles cooperate reliably: how messages are scheduled, how warnings travel and how sensors share a common sense of time.
At NVIDIA, I work in systems software. My research connects 5G NR sidelink and aerial–terrestrial communication with dependable real-time systems, including Precision Time Protocol (PTP) and gPTP.
“Getting one robotaxi to drive is hard. Proving that many can operate safely, every day, is a different kind of challenge.”
For introductions, collaborations and conversations about the systems behind AI.
Anwesha Kar is a Systems Software Engineer at NVIDIA working on intelligent transportation systems (ITS), vehicle-to-everything (V2X) communication and real-time synchronisation using the Precision Time Protocol (PTP). Her research spans 5G NR sidelink scheduling, aerial–terrestrial cooperation and dependable timing for connected and autonomous mobility.
Download bio ↓The parts that let it work reliably: timing, communication, validation and clear limits on what it can do. When software acts in the physical world, those details become part of the outcome.