# Sharadh Ramaswamy

> Engineering Leader, Applied Researcher at Meta Superintelligence Labs (MSL)

Location: Sunnyvale, California, United States
Profile: https://flows.cv/sharadh

Audio, video understanding at MSL

## Work Experience
### Staff Software Engineer @ Meta
Jan 2020 – Present | United States
Tech Lead for
-  Quadrillion Token Multimodal Datasets
-  Large Scale Inference Systems
-  Video Understanding AI

### Software Engineer @ Google
Jan 2014 – Jan 2020 | Mountain View, CA
Tech Lead in Geo Machine Perception (Google Maps)

Tech Lead in Machine Intelligence + Computer Vision (Google Photos)

Products : PhotoScan, Google Lens and other projects

### Research Scientist (Computer Vision ) @ Lab126
Jan 2013 – Jan 2014 | Sunnyvale
- Dynamic perspective for the Amazon phone!
- Head tracking algorithms, false positive rejection, failure analysis
- Factory and online stereo calibration for the Amazon phone
- Built a map-reduce device farm for large scale CV evaluation
- Staffed a world class CV team building breakthrough product features
- Developed prototype product features, demoed with Android apps
- 20+ published patents (USPTO Link: https://goo.gl/s9SKd5)

### Software Engineer, Emerging Technology @ Lab126
Jan 2012 – Jan 2013 | Sunnyvale, CA
- Evaluated emerging embedded technologies and generated IP
- Built computer vision systems for Amazon's digital products

### Research Staff @ Mayachitra, Inc.
Jan 2009 – Jan 2012
Summary: I built advanced video/image analysis and database management systems. 

Specifics:
- Led a team of researchers in the development of state-of-the-art activity descriptors for aerial video streams (used by DARPA for surveillance)
- Developed business plans and pursued DARPA solicitations for research and product development at Mayachitra
- Developed hitherto fastest exact similarity index and novel clusterings for activity descriptors,
- Designed constant time video/image median filters 
- Developed multi-threaded (with TBB) SIMD C++ libraries for activity description, K-means clustering and similarity indexing of large image/video feature data-sets.

Programming/Software Skills developed/used: 
C/C++, Python, Matlab,  Object oriented design principles (OOP), Unit-testing, Version Control (SVN/Git), GUI design, Multi-threaded design, Cross-platform build

Packages used : 
Boost, OpenCV, Intel Integrated Performance Primitives (IPP), Intel Thread Building Blocks (TBB), Qt4, Standard Template Library (STL), HDF5, CMake, Visual Studio, Eclipse, XCode

### Graduate Student Researcher @ Signal Compression Lab, ECE Dept., University of California, Santa Barbara
Jan 2003 – Jan 2008
Summary: I researched novel applications of signal compression in large-scale similar image/video search (indexing), multi-data-stream retrieval (fusion coding) and distributed compression in large sensor networks.

Specifics:

Fusion Coding
- Developed the Fusion Coder (FC), to optimize the storage space-retrieval time-quality tradeoff in multi-data-stream processing; FC provided enormous gains over naive approaches (joint compression (VQ) and scalar quantization).

- Scaled FC designs to large stream (source) systems by exploiting distributed and predictive computing architectures. 

Similarity Search/Indexing
- Designed novel indexing schemes (based on clustering/VQ ) for scalable similarity search in very large, high-dimensional feature databases (shown to outperform state-of-the-art systems by factors reaching 100X) .

- Designed indexing schemes to allow for user relevance feedback wherein the (Mahalanobis) distance measure changes periodically (shown to outperform state-of-the-art systems by factors reaching 1000X). 

Distributed Compression
- Designed error-resilient, distributed compression systems for very large sensor networks (opening new vistas for real world deployment of sensor networks).

### Intern @ Invention Labs (India) Inc.
Jan 2007 – Jan 2007
Designed systems for authentication/validation of Indian currency notes in automated teller machines. Applied novel signal processing concepts for feature extraction and recognition. Several features were successfully identified including watermark, security thread, latent image.

### Intern @ Qualcomm
Jan 2006 – Jan 2006
I studied the effects of receiver automatic gain control (AGC) misalignment on WCDMA (HSDPA/DCH) channel frame error rates, pilot channel (CPICH) SNR estimates (for CQI) and the benefits of feed-forward vs. feedback AGC implementations.


## Education
### Ph.D in Signal Compression, Search and Retrieval, Sensor Networks
UC Santa Barbara

### M.Tech in Communications Engineering
Indian Institute of Technology, Madras

### B.Tech in Electrical Engineering
Indian Institute of Technology, Madras

### Padma Seshadhri Bala Bhavan


## Contact & Social
- LinkedIn: https://linkedin.com/in/rsharadh
- Portfolio: http://sites.google.com/site/sharadhramaswamy/

---
Source: https://flows.cv/sharadh
JSON Resume: https://flows.cv/sharadh/resume.json
Last updated: 2026-04-12