CURRICULUM VITAEREV 2026.07

Rajasekhar Josyula

Ten years of perception, from photons to planners.

01 // EXPERIENCE

  1. Staff Autopilot Software Engineer_

    2022.05 — present

    Tesla AI · Palo Alto, CA

    • Train and ship multi-task vision models for driver monitoring and cabin monitoring: drowsiness detection, gaze and head-pose estimation, attentiveness and distraction classification, and cabin-state understanding.
    • Own the DMS data engine end to end — fleet-scale data mining, auto-labeling, active learning, and hard-example curation feeding continual retraining and shadow-mode evaluation.
    • Deliver European homologation features (EU GSR DDAW / ADDW) from dataset design and operating-point selection through on-device INT8 deployment and regulatory validation.
    • Design and optimize sub-10ms photon-to-NN camera pipelines for production cars and Optimus robots; drive camera integration for Cybercab, Robotaxi, and next-gen platforms.
    • Evolve real-time camera stacks in C++20 with strict determinism, and ensure Robotaxi stability through fault-tolerant frame routing and production stress testing.

    MULTI-TASK LEARNING · DMS / DDAW · ACTIVE LEARNING · TENSORRT / INT8 · C++20 · CUDA

  2. Sr. Software Engineer, Autopilot_

    2020.02 — 2022.05

    Tesla · Palo Alto, CA

    • Led camera bring-up and Sony IMX sensor validation for Model 3/Y/S/X platforms.
    • Developed vehicle-level image quality pipelines covering MTF, dynamic range, and color accuracy.
    • Maintained production C++ camera capture pipeline (Bayer → CCM → gamma → tone mapping).
    • Drove closed-loop IQ tuning using fleet data; deployed OTA updates across the global fleet.

    IMAGE PROCESSING · SONY IMX · FLEET OTA · TENSORRT

  3. Sensor Software Engineer, Autopilot_

    2017.05 — 2020.02

    Tesla · Palo Alto, CA

    • Led factory camera calibration for vision-only Autopilot using ChArUco diamond boards.
    • Optimized embedded ISP pipelines for rapid auto-exposure and white-balance convergence.
    • Owned camera intrinsic generation at contract manufacturers with temperature-swept captures.

    CALIBRATION · ISP · COMPUTER VISION · MANUFACTURING

  4. Software Engineer, Quality_

    2016.01 — 2017.05

    Tesla · Fremont, CA

    • Drove product improvement for Model X using Python, Tableau, and Flask-based tools.
    • Performed root-cause analysis on authentication failures; implemented OTA countermeasures.
    • Built data pipelines correlating telemetry with Consumer Reports and JD Power metrics.

    PYTHON · DATA ANALYSIS · QUALITY · TELEMETRY

02 // OBJECTIVE

System Objective: Minimize the generalization gap between offline model training and real-world edge scenarios. Optimize the Pareto frontier of hardware efficiency (sub-10ms latency) and high-recall safety metrics, shipping models that map raw photons to deterministic control actions.
𝓛(θ) · focal loss · cosine LRearly stopepochs →trainvalprecision–recalloperating pointrecall →P
Fig. 2 — Pareto optimization loop. Scrub the validation loss curve to verify convergence and prevent overfitting; scrub the precision–recall frontier to establish the operational safety margin where neural confidence meets regulatory and physical safety limits.

03 // RESEARCH_THREADS

Occupancy networksBEV transformersVision-only perceptionEnd-to-end planningTemporal fusionGaze estimationDrowsiness modeling (KSS)Uncertainty calibrationKnowledge distillationQuantization-aware trainingFleet / shadow-mode evalAuto-labeling

04 // INSTRUMENTATION

C++ / C
95
Python · PyTorch
90
TensorRT · INT8 / PTQ · ONNX
85
Data engine · active learning
85
TypeScript
80
CUDA · Triton kernels
75
Rust
75
Distributed training · DDP
70
RTOS · QNX · Zephyr
60

hover an instrument to see where it is used

05 // TRAINING

M.S. Information Systems & Security
University of the Cumberlands · GPA 4.0 · Cybersecurity focus
2017
M.S. Electrical & Electronics Engineering
Lamar University · GPA 3.6 · Embedded systems focus
2016
B.E. Electrical Engineering
Amrita University · Robotics focus
2013