diff --git a/resume.tex b/resume.tex
index e0473bb..17e9d96 100644
--- a/resume.tex
+++ b/resume.tex
@@ -108,21 +108,19 @@
       \resumeItemListEnd
   \resumeSubHeadingListEnd
 
-\begin{comment}
 %-----------EXPERIENCE-----------------
 \section{Work Experience}
   \resumeSubHeadingListStart
     \resumeSubheading
-      {Family Investment Office}{Plano, TX}
-      {Quantitative Analyst}{Q4 2024 -- Present}
+      {CoGuide (\href{https://coguide.ai}{coguide.ai})}{Austin, TX / Irving, TX / San Francisco, CA}
+      {Founding Engineer / Member of Technical Staff}{Jan 2026 -- Present}
       \resumeItemListStart
-        \resumeItemNH{Engineered and executed medium-term volatility trading strategies across equity options and index derivatives, resulting in a 60\% portfolio return over one quarter.}
-        \resumeItemNH{Developed predictive models leveraging implied-volatility surfaces, realized variance, and skew metrics to identify dislocations and structure trades.}
-        \resumeItemNH{Built custom infrastructure in Rust and Python for backtesting, signal generation, risk metrics, and position management.}
-        \resumeItemNH{Managed exposure via delta-hedged structures, calendar spreads, and straddle/strangle plays, with automated alerts.}
+        \resumeItemNH{Build systems for an AI instructional-insight platform that turns real classroom audio into private coaching feedback for teachers and aggregate engagement trends for schools.}
+        \resumeItemNH{Designed and built the entire production backend in Rust ($\sim$30k-line codebase, closed source) and stood up the full AWS infrastructure, scaling the platform from zero product to deployments in 10+ schools.}
+        \resumeItemNH{Engineered the audio transcription and speaker-diarization pipeline with custom CUDA fusion kernels and VAD-based batching on legacy NVIDIA T4 GPUs --- 2x the cost efficiency of market alternatives and 10x that of AWS Transcribe.}
+        \resumeItemNH{Engineered a two-stage computer-vision pipeline (YOLOv8 person detection + upper-body pose estimation) detecting hand-raise engagement events in classroom video at $\sim$0.96x real-time on GPU, emitting annotated video and timestamped JSON/NDJSON event logs --- \href{https://git.cyber.ayyalasomayajula.net/marsultor/CoGuide-Miniproj}{open-source miniproject}; production work is closed source.}
       \resumeItemListEnd
   \resumeSubHeadingListEnd
-\end{comment}
 
 
 %-----------PROJECTS-----------------