|
1 | 1 | ## Senior Software Engineer, Graphics Software Infrastructure |
2 | | -### [Qualcomm](https://www.qualcomm.com/snapdragon/overview) — San Diego, CA |
3 | | -#### 2021 - Present |
4 | | -##### *Boulder · Toronto · San Diego - Promoted 2024* |
5 | | - |
6 | | -- `Architected and own Aether E2E`: an in-house `end-to-end automation platform` purpose-built to orchestrate pipelines dispatching `200K+ jobs/week` as containerized (`Docker`/`Singularity`) workloads across an `on-prem LSF server-farm`, with real-time result reporting for `Vulkan`/`OpenCL`/`OpenGL` graphics driver validation across all upcoming premium-tier flagship `[Snapdragon Adreno GPUs](https://www.qualcomm.com/processors/adreno)`. |
7 | | - |
8 | | -- `Engineered FastAPI backend` + `RedisJSON`/`RediSearch` for `sub-second` live access to job statuses, test results, and aggregated result metrics; sustains `~2M rolling rows/month` with custom expiry model, deliberately trading history depth for query performance at scale. |
9 | 2 |
|
10 | | -- `Decoupled` test execution from reporting by routing results via messaging queues with `RabbitMQ` + `Postgres` database for tracking regressions and sending test logs to `Artifactory` for persistent storage as they finish, enabling `independent scaling of each layer`. |
11 | | - |
12 | | -- `Designed delightful user experiences` through interactive `Web dashboards`, `Excel spreadsheets` to show up-to-date result trends, jobs health and test regressions, enabling stakeholders for quick decision making. |
| 3 | +### [Qualcomm](https://www.qualcomm.com/snapdragon/overview) — San Diego, CA |
13 | 4 |
|
14 | | -- `Established AeResult`, a `Pydantic`-based canonical result schema unifying data contracts across Aether's pipeline, extensible enough that the on-device `benchmarking team adopted it wholesale` for KPI workloads without platform modifications, `saving months of parallel engineering effort`. |
| 5 | +#### 2021 - Present |
15 | 6 |
|
16 | | -- `Introduced MCP servers` wrapping live `API` backends, making operational infra `LLM`-queryable via natural language without direct API knowledge; presented org-wide to drive `MCP` and agentic coding (`Cline`) adoption across engineering teams. |
| 7 | +- `Architected and own` an in-house pre-silicon `GPU validation` `end-to-end automation framework` dispatching `500K+ jobs/week` on `LSF` on-prem, catching driver regressions on flagship `Snapdragon Adreno GPUs`. |
| 8 | +- Parallelized `Docker`/`Singularity` containerized workloads, cutting nightly driver validation cycles from `240 days to 1 day` to support higher throughput. |
| 9 | +- Delivered a `FastAPI` + `RedisJSON`/`RediSearch` backend with `TTL`-based expiry, providing `sub-second` access to live test metrics with `2M+ rows/month`. |
| 10 | +- Designed `web dashboards`, on-demand `Excel` reports, and automated regression `email alerts` to `meet each stakeholder group's reporting needs`. |
| 11 | +- Eliminated reporting bottlenecks via async `RabbitMQ` routing to `Postgres` and `Artifactory`, enabling independent scaling of each layer. |
| 12 | +- Standardized `Pydantic` canonical schemas across `10+` Khronos test suites; adopted by the benchmarking team, `saving months of engineering effort`. |
| 13 | +- Introduced `MCP servers` wrapping live `API` backends, making `LLM-queryable operational infra`; drove org-wide adoption of `agentic coding` (`Cline`). |
17 | 14 |
|
18 | 15 | --- |
19 | 16 |
|
20 | 17 | ## Data Science Developer, Applied ML |
21 | | -### [KORE Wireless](https://www.korewireless.com/) — Blue Ash, OH |
22 | | -#### 2020 - 2021 |
23 | | -##### *R&D, Applied ML, Predictive Analytics, Clustering, Anomaly Detection; PySpark, Scikit-Learn, FastAPI, PowerBI, Jupyter Notebooks* |
24 | 18 |
|
25 | | -- `Engineered anomaly detection APIs` (`Auto-Thresholds`, `Auto-Device Grouping`) and `ETL pipelines` processing live `IoT` telemetry from `NetFlow`/`Radius` sources to power the `[SecurityPro](https://www.korewireless.com/news/koreone-security-connectivity-analytics-platform-enable-innovative-i/)` network diagnostic product. |
| 19 | +### [KORE Wireless](https://www.korewireless.com/) — Blue Ash, OH |
26 | 20 |
|
27 | | -- `Owned the full ML lifecycle (0-to-1)` from rapid prototyping through deployment; built `PowerBI` dashboards with `3D clustering` visualizations to communicate model outputs; presented weekly progress directly to `C-suite` (`CTO` & `SVPs`), translating ML outcomes into business decisions. |
| 21 | +#### 2020 - 2021 |
28 | 22 |
|
29 | | -- Converted to full-time following successful delivery of core anomaly detection features in Winter `2020`. |
| 23 | +- `Engineered anomaly detection APIs` (`Auto-Thresholds`, `Auto-Device Grouping`) and `ETL pipelines` processing live `IoT` telemetry from `NetFlow`/`Radius` sources to power the `SecurityPro` network diagnostic product. |
| 24 | +- `Owned the full ML lifecycle (0-to-1)` from prototyping through deployment; built `PowerBI` dashboards with `3D clustering` visualizations; presented weekly to `C-suite` (`CTO` & `SVPs`), translating ML outcomes into business decisions. |
30 | 25 |
|
31 | 26 | --- |
32 | 27 |
|
33 | 28 | ## Associate Researcher, Deep Learning |
34 | | -### [Marine Bio-acoustics Research Collaborative (<u>MBARCO</u>)](https://acoustics.ucsd.edu) — San Diego, CA |
35 | | -### [Marine Acoustics Research (<u>MAR</u>) Lab, SDSU](https://roch.sdsu.edu/index.php/research-overview) & [Scripps Institution of Oceanography (<u>SIO</u>), UCSD](https://www.cetus.ucsd.edu) |
36 | | -#### 2019 - 2020 |
37 | | -### Master's Thesis: `[Learning to Detect Odontocete Whistles from Generative Synthetic Samples](https://csu-sdsu.primo.exlibrisgroup.com/permalink/01CALS_SDL/10r4g1c/cdi_proquest_journals_2493456813)`, Advisor: [Dr. Marie Roch](https://roch.sdsu.edu) |
38 | | -##### *Unsupervised Learning, Bio-Acoustics, Computer Vision, Speech Processing, CycleGAN, WGAN, CNN, ResNet, UNet; Python, PyTorch, CUDA* |
39 | 29 |
|
40 | | -- Designed **[DeepWhistleGAN](https://drive.google.com/file/d/1vp1WcMvt0eAPQbaeXxriQ5qnka3edRHn/view?usp=sharing)**, a novel hybrid architecture addressing data scarcity in marine bio-acoustics; with `10x synthetic data augmentation` via `WGAN`+`CycleGAN` enabling model convergence on just `6.25% of annotated data`. Achieved `80.5% F1`, `96.6% Precision`, outperforming baselines by `8%`. |
| 30 | +### [Marine Bio-acoustics Research Collaborative (MBARC)](https://acoustics.ucsd.edu), [SDSU](https://roch.sdsu.edu/index.php/research-overview) — San Diego, CA |
41 | 31 |
|
42 | | -- `Proposed transfer learning` pipeline to scale `PAM (Passive Acoustic Monitoring)` across endangered whale species for `marine mammal conservation`. |
43 | | - |
44 | | -- Presented research findings to `ONR (Office of Naval Research)` federal research sponsors on model performance and conservation impact. |
45 | | - |
46 | | ---- |
| 32 | +#### 2019 - 2020 |
47 | 33 |
|
48 | | -## Research Intern, Recommendation Systems |
49 | | -### HireValley Inc — Ahmedabad, India |
50 | | -#### 2016 - 2017 |
51 | | -##### *Recommendation System, NLP, Microservices, Ontology; Python, Flask, SPARQL, RDFLib* |
| 34 | +### Master's Thesis: `[Learning to Detect Odontocete Whistles from Generative Synthetic Samples](https://csu-sdsu.primo.exlibrisgroup.com/permalink/01CALS_SDL/10r4g1c/cdi_proquest_journals_2493456813)`, Advisor: Dr. Marie Roch |
52 | 35 |
|
53 | | -- Co-developed a `[job-candidate recommendation engine](https://drive.google.com/drive/folders/0B9gQb-9dKj0uM1h5eWRlRXhpUmc?resourcekey=0--sNi2HxA2uTPENpWuYTB-A&usp=sharing)` on a cloud-based `AWS EC2` `microservice` platform; built `NLP` feature extraction pipelines from resume and job descriptions to build a skill-based ontology using `SPARQL`/`RDFLib` for knowledge inference. `[Published](https://ieeexplore.ieee.org/document/8369531/)` at `IEEE SysCon 2018`. |
| 36 | +- Designed a novel hybrid architecture addressing data scarcity in marine bio-acoustics; `10x synthetic data augmentation` via `WGAN`+`CycleGAN` enabling model convergence on just `6.25%` of annotated data. Achieved `80.5% F1`, `96.6% Precision`, outperforming baselines by `8%`. |
| 37 | +- Proposed `transfer learning` pipeline to scale `PAM (Passive Acoustic Monitoring)` across endangered whale species for `marine mammal conservation`; presented model performance and conservation impact to `ONR (Office of Naval Research)` federal research sponsors. |
0 commit comments