Srihan Balaji

I'm building real-time signal processing for brain-computer interfaces as a Research Engineer at the Systems Neural Engineering Lab at Emory and Georgia Tech, whose BCI work is part of the BrainGate clinical trial.

My engine turns 30 kHz recordings from up to 2,048 channels into spike features for speech and neural-to-text decoding, now 6.3× faster with zero bin mismatches.

Before that, I studied dopamine release with fast-scan cyclic voltammetry as an NIH IRTA fellow, built ML systems at Lockheed Martin and AWS, and went through Y Combinator at 19 with Protent Labs, a real-time video analytics company I founded.

I study Electrical Engineering and Computer Science at Georgia Tech, concentrating in bioengineering and signal processing. Here's my resume.

Experience

  1. Systems Neural Engineering Lab · Research Engineer 2026 – now

    Real-time spike extraction from 30 kHz, 2,048-channel neural recordings for BCI speech and neural-to-text decoding.

    Details
    • Engineered a real-time signal processing engine that ingests 30 kHz neural recordings across up to 2,048 channels and extracts spike patterns for downstream BCI applications.
    • Replaced per-millisecond matrix reconstruction in Common Average Referencing with a cached block projection matrix (I − P), and turned acausal 250–5,000 Hz FIR bandpass filtering into a single-pass matrix-vector multiply.
    • Cut processing wall time per second of neural data from 14.08 s to 2.25 s at 2,048 channels (4.09 s to 1.09 s at 1,024), with zero bin mismatches across 25,600 samples, streaming sparse CSR and event outputs.
  2. Protent Labs · Founder & CEO (technical), Y Combinator 2025 – now

    Real-time multi-camera video analytics deployed on-site with law enforcement and municipal IT; raised $500K.

    Details
    • Built the end-to-end production stack for multi-camera ingest and GPU-scheduled model serving over RTSP, WebRTC, and Low-Latency HLS, integrating with customers' existing VMS architectures.
    • Shipped one codebase into diverse environments, including air-gapped on-prem GPU installs with local secrets and licensing.
    • Cleared agency security reviews and procurement as the sole technical counterpart (CJIS-aware data handling, SSO, audit logging).
    • Filed a provisional patent for real-time video anomaly detection and directed 5 engineers from Georgia Tech, Caltech, UVA, and Virginia Tech.
  3. Lockheed Martin · Research Engineering Intern, ATL 2025

    PyTorch training-and-inference pipeline under a sub-10 ms latency budget; cited in a winning $18M DARPA proposal.

    Details
    • Framed batch scheduling as constrained optimization to raise throughput 35% inside a sub-10 ms latency budget; the prototype was cited as proof-of-concept in a winning $18M DARPA proposal.
    • Developed a forecasting engine (EWMA, ARIMA, hypothesis testing) for runtime latency and hardware utilization; proactive control limits cut variance 42%.
    • Wrote an offline back-testing framework with block-bootstrap resampling and stress tests to quantify tail risk before deployment.
  4. Amazon Web Services · Software Development Engineer Intern 2025

    Spatio-temporal CNNs over multi-channel sensor streams on SageMaker; automated testing for a 100+ channel SDR rack.

    Details
    • Designed spatio-temporal convolutional networks to classify multi-channel sensor streams, training on 2.5M+ samples per day, with production training and inference on SageMaker over terabytes of streaming telemetry.
    • Integrated GNU Radio into a software/hardware acquisition framework spanning hundreds of concurrent RF channels, removing ~15 hours/week of manual testing.
    • Automated bring-up and regression validation of a 100+ channel SDR rack (Ethernet, SPI, I2C), cutting full-rack test time from 12 hours to 90 minutes.
  5. National Institutes of Health · Research Intern (IRTA Fellow) 2023 – 2024

    Fast-scan cyclic voltammetry of phasic dopamine release in the Laboratory for Integrative Neuroscience.

    Details
    • Ran fast-scan cyclic voltammetry in dorsolateral striatum and nucleus accumbens to measure how adenosine antagonists change phasic dopamine release.
    • Wrote MATLAB and Python analysis code to extract dopamine transients and compare timing and amplitude across drug conditions and recording sites.
    • Selected as an NIH Intramural Research Training Award fellow and chosen to present at the 2024 NIH Summer Poster Day.

Research

  1. Voltammetric Analysis of Dorsal Striatal Dopamine Dynamics Following Chronic Ethanol Treatment in C57BL/6J Mice

    NIH · NIAAA Laboratory for Integrative Neuroscience · with V. Mamillapalli, S. Badaro, Y. Mateo, D. M. Lovinger

    Chronic ethanol exposure caused a significantly larger drop in dopamine release in the dorsolateral (habit) striatum than the dorsomedial (goal-directed) striatum, consistent with a shift toward habitual behavior in alcohol use disorder.

    View · PDF

  2. CLOUD-D RF: Cloud-based Distributed Radio Frequency Spectrum Sensing

    AWS · Virginia Tech Hume Center · 2025 National Security Colloquium · with A. Dillard, R. Ghaleb, A. Pappu, R. Thaboun, D. Green, A. Jones, M. Toubeh, C. Headley

    Per-radio modulation classifiers trained on GNU Radio datasets, with learned features streamed over ZeroMQ to a central fusion server (AWS as the planned fusion center) for collaborative decisions.

    View · PDF

Education

  1. Georgia Institute of Technology · B.S. EE, B.S. CS2028

    Concentrations in Bioengineering and Signal & Information Processing. GPA 4.00.

  2. Y Combinator2026

    Raised $500K for Protent Labs at 19.

  3. Thomas Jefferson High School for Science and Technology2024

    GPA 4.00.

Languages Python, C++, Rust, C, Go, SQL, Java, JavaScript, MATLAB
ML & systems PyTorch, OpenCV, AWS SageMaker, Linux, WebRTC, RTSP, LL-HLS
Signals & hardware DSP, GNU Radio, FPGAs, ARM microcontrollers, KiCad, Quartus II, LabVIEW

Moments

Two Protent Labs founders on stage at a Y Combinator event, one speaking into a microphone in front of an audience.
Presenting Protent Labs at Y Combinator.
Four friends in matching white Miles for Minds Fall 2025 JED shirts, standing arm in arm outdoors among autumn trees.
Miles for Minds, Fall 2025, a run supporting the JED Foundation's mental-health work.
Download