Arnav Angarkar
Open to opportunities · India or remote
I make things fast: live translation, drone ground control, and ML on tiny devices. ⚡
CS undergrad at IIIT Dharwad (minor in Quantum Computing, class of 2028). Google Summer of Code 2026 contributor and Eventyay maintainer at FOSSASIA. Previously ML at IISc and Poonawalla Fincorp, and I led our team to All India Rank 3 in ISRO's IROC Grand Challenge.
- ISRO IROC
- AIR 3
- Led our team through the autonomous UAV challenge.
- FOSSASIA
- Open-source maintainer
- VoxBento & Eventyay.
- Poonawalla Fincorp
- ML in production
- Built an in-house alternative to a ~₹2 Cr/year vendor service.
- National hackathons
- 5 wins & podiums
- From offline AI to hardware and developer tools.
Experience
May 2026 – Present
GSoC Contributor, VoxBento · Google Summer of Code · FOSSASIA
Remote
- Architected VoxBento from scratch: a low-latency, browser-based multilingual interpretation platform for conference-scale events.
- Live audio over WebRTC (WHIP/WHEP) through MediaMTX, kept out of Python; FastAPI coordinates booths and async speech-to-text, translation and TTS workers over WebSockets.
- Presented it at FOSSASIA Summit 2026, demoed it live at Wikimania Paris 2026, and wrote up the whole journey on this site.
WebRTCMediaMTXFastAPIPostgreSQLOct 2025 – Present
Project Maintainer, Eventyay · FOSSASIA
Remote
- Maintain Eventyay, a 6k+ star open-source conference platform used by Wikimedia and PyCon.
- Authored 221 merged PRs in Eventyay across ticketing, scheduling, speaker management and platform infrastructure; review contributions and keep production stable.
DjangoJinjaVue.jsLinuxJun 2025 – Dec 2025
ML Systems Intern · IISc · RIMM Lab
Bangalore
- Built a modular inference system combining MediaPipe landmark detection with segmentation-based garment synthesis for full-body virtual try-on.
- Privacy-first inference APIs for edge devices; cut latency through quantization and on-device processing.
PyTorchMediaPipeQuantizationJun 2025 – Jul 2025
ML/AI Intern · Poonawalla Fincorp
Pune
- Deployed a self-hosted, distributed ML inference pipeline for identity verification on AWS, replacing a vendor solution valued at ~₹2 Cr/year.
- Built self-hosted vector search with sub-50 ms nearest-neighbour latency at production scale, integrated into fraud-prevention workflows.
AWSVector searchPythonJun 2025 – Aug 2025
Project Intern & Team Lead · QWorld · QIntern 2025
Remote
- Led a team of 7 building QPlay, a game-based platform for learning quantum computing (Bell states, CHSH, decoherence).
Quantum computingLeadershipNov 2024 – Aug 2025
Systems Engineer & Team Lead · ISRO IROC Grand Challenge
India
- Led the team to All India Rank 3 of 2,000+ teams in a 10-month autonomous UAV challenge.
- Built a custom MAVLink ground control system with real-time telemetry and 3D flight visualization; sensor fusion and SLAM on a Jetson Nano with Pixhawk, LiDAR and depth cameras.
- Computer vision for safe landing zone detection (YOLOv5, Depth Anything, DINOv2), optimized for embedded hardware.
C/C++MAVLinkJetsonSLAMDec 2024 – Mar 2025
ML Intern · IISc · CPDMed TBI (Aarnaq)
Bangalore
- Fine-tuned YOLOv8 for number-plate recognition on urban mobility data; optimized with ONNX, FP16 and TensorRT for real-time embedded deployment.
YOLOv8ONNXTensorRTNov 2024 – Apr 2025
Research Intern · Brahmand Exploration
Pune
- Built an NLP chatbot over the company’s satellite-development knowledge base and revamped brahmandexploration.com into a digital encyclopedia.
NLPWeb
Selected projects
VoxBento
GSoC 2026
Low-latency, browser-based multilingual interpretation for virtual and hybrid events. Built from scratch for GSoC 2026.
DataMind
1st · ZeroDay MLH
Natural-language data agent where every answer carries its evidence: the executed SQL, bound parameters, result rows and Snowflake query ID.
RagVault
Runner-up · HAL Aerothon '26
Fully offline, privacy-first enterprise RAG system. Document intelligence with zero data leaving the network.
Transmon-QKD-ML
Research · IIIT Dharwad & IIT Delhi
How well does vendor calibration data predict BB84 error rates? A 25-day hardware study on a 156-qubit IBM Heron processor.
CerebrumAI
Offline-first multimodal healthcare triage engine with quantized LLMs, deployable on a Jetson Nano.
Telemetry Power Governor
1st · HACKMAZE '26
Configurable Verilog RTL IP for telemetry-driven dynamic power management in SoCs.
More projects
- 2025CampusDabbaPre-incubated campus marketplace startup I founded. Raised 2,500 USD in seed funding.
- 2024Customer Service Conversation AnalyzerReal-time call analysis: speech-to-text, sentiment and automated SOP compliance checks.
- 2024AutoML-MLOps Platform PrototypeEnd-to-end AutoML prototype with experiment tracking, orchestration and an LLM guidance bot.
- 2024Dynamic API for NoSQL QueryBackend API for single and bulk MongoDB upserts with flexible filters.
- 2024Binary Classification Web AppStreamlit app comparing SVM, logistic regression and random forest classifiers on any dataset.
- 2024BMP Image FilterCommand-line tool in C applying grayscale, sepia, reflection and blur filters.
Talks & events
Nov 2026upcoming
OpenTechSummit India 2026
Mumbai · at Devcon 8 · Speaker · 3 sessions
Aug 2026
Opportunity Open Source Conference 4.0
IIIT Allahabad · Speaker
Jul 2026
Wikimania 2026
Paris · Presenter
- “Live demo of VoxBento and Eventyay”
Mar 2026
FOSSASIA Summit 2026
Bangkok · Speaker
- “Talk on Eventyay and real-time multilingual streaming”
Research
- Predicting BB84 Error Rates from Transmon Calibration Data: Learned Surrogates versus Physical Baselines on a 156-Qubit IBM Heron Processor
with Aswath Babu (IIIT Dharwad) · supervised by Aswath Babu and Prof. V. Ravishankar (IIT Delhi) · Manuscript, 2026
25 days, 115 calibration snapshots and 105 BB84 runs on ibm_marrakesh: readout error alone predicts QBER better than ML surrogates (Spearman ρ 0.61 vs 0.47).
- Temporal Graph Neural Models for Predicting Organizational Silos in Evolving Communication Networks
Under review, 2026
- Quantum Error Correction: Understanding from Bell States to Surface Codes
ResearchGate, 2025
Awards & recognition
- All India Rank 3, ISRO IROC Grand Challenge · Top 3 of 2,000+ teams2025
- 1st Place, MLH ZeroDay Hackathon · DataMind · Open Source Track2026
- 1st Place, HACKMAZE '26 · Telemetry Power Governor RTL2026
- 4th Place, UIDAI BioChallenge (Face Match) · 0.993 AUC, age-invariant face matching2026
- Runner-up, HAL Aerothon '26 · RagVault2026
- 1st Place, DevHack '25, IIT Dharwad · CampusDabba, 500+ applicants2025
- Winner, Hack2Future · Only freshman team in the final, 400+ applicants2024
- Google Summer of Code Contributor · FOSSASIA, one of ~1,200 selected globally2026
- Linux Foundation Scholar
- Bug bounty acknowledgements · Netflix, Google, 1Password, Cursor, Udemy2022 – 2025
Writing
My GSoC 2026 journey
VoxBento’s architecture, every FOSSASIA PR, the final report and photos from Bangkok, Paris and Germany.
- Apr 18, 2025Hero’s Journey: The Necessary Darkness, Basketball, and RebuildingPersonal Reflections
- May 1, 2026Community Bonding: The Architectural Epiphany of Keeping Python Out of Live AudioGSoC 2026
- May 9, 2021Me In Next Five Years? (Written in 10th Grade, and How It Came True)School Days & Reflections
- Mar 3, 2026FOSSASIA Summit 2026: Bangkok, Street Food, and the Spark for VoxBentoGSoC 2026
Education
IIIT Dharwad
B.Tech, Computer Science & Engineering · Minor in Quantum Computing
2024 – 2028 · CGPA 8.81 · Quantum minor 10.00
Skills
- Languages
- Python, C/C++, TypeScript, SQL, Verilog
- Systems & Backend
- FastAPI, Django, Redis, PostgreSQL, WebRTC, WebSockets, Linux
- ML & AI
- PyTorch, ONNX, TensorRT, LangChain, Llama.cpp, RAG, Computer vision
- Infra & Tools
- AWS, Docker, GitHub Actions, MLflow, Snowflake, Next.js
- Quantum
- Qiskit, Cirq, QEC, BB84 QKD
Contact
Looking for a backend, systems or ML engineer? I’d love to talk.
Based in Pune, India. Open to roles anywhere in India or remote.