MISSION.INITIALIZATION(AUTHENTICATING)
Establishing secure uplink to neural starfield...
System.Initialize(Visionary)

Vedang Avaghade

M.S. CS @ ASU  |  AI/ML Engineer

Bridging the gap between theoretical AI research and production-grade software. I build systems that think, reason, and scale (usually without catching fire).

class Agent(LLM): def align_to_reality(self): while self.is_hallucinating: self.ground_truth(reality) # TODO: Lower temperature try: return self.innovate(problem) except DistributionShift: self.backpropagate(loss=Reality) # TODO: Fix inference latency
Scroll

Origin Story

Technology isn't just about code; it's about mitigating chaos.

My journey began with a fascination for how data can shape the physical world. This drove me to C-DAC, where I didn't just build apps (I mostly fought fires). I helped modernize India's rural infrastructure monitoring, touching the lives of millions through the 'Quality First' & 'Meri Sadak' initiative. (Which, let's be honest, mostly involved explaining to stakeholders why "cloud" didn't mean actual clouds.)

Currently, I am diving deeper into the "brain" of modern computing at Arizona State University. My focus is on the subtle intricacies of Large Language Models.

System.Architecture AI.Research Runtime.Optimization
SYST_INIT_0xFF12 // CORE_LOAD // AGENT_READY
TECH_STACK // EDU_CORE // KNOWLEDGE_BASE

Education

M.S. in Computer Science
Arizona State University (Tempe, AZ)
Aug 2025 - Present

Advanced coursework in Agentic AI, Natural Language Processing, Software Development, and Security. (Mastering the architectural constraints of scaling intelligence across decentralized nodes.)

Natural Language Processing Agentic AI Foundations of Algorithms Statistical Machine Learning Data Processing at Scale
ASU // KNOWLEDGE_GRAPH
B.Tech in Computer Science & Engineering
MIT World Peace University (Pune, India)
Jul 2019 - Jun 2023

Comprehensive curriculum covering the fundamentals of Computer Science and Engineering. (Building the foundational matrices for structural logic and gradient optimization.)

Object Oriented Programming Data Structures Operating Systems Database Management Systems Artificial Intelligence Machine Learning
MIT WPU // CORE_FOUNDATIONS
MISSION_LOGS // EXEC_PROT // OPER_XP

Experience

Project Engineer (Rural Tech)
Center for Development of Advanced Computing (C-DAC, India)
Nov 2024 - Jul 2025

The problem? Ensuring data flowed reliably from remote villages where signal bars were hallucinations. By managing real-time GPS tracking and analytics for 98,500 km of rural roads, we had to cut latency without reliable networks. (No pressure, right? Just the connectivity of an entire subcontinent riding on my git commits.)

My insight? Don't trust the network. I directed full-stack engineering (Flutter, .NET Core, MySQL) with an aggressive offline-first REST API and JWT authentication. It allowed 150,000+ users to sync their work seamlessly, reducing field reporting latency by 40%.

C-DAC // LATENCY_SWEEP
Software Engineer (Backend)
Jombay, Pune, India
Oct 2023 - Apr 2024

I walked into a startup where the messaging infrastructure was a house of cards. I re-architected the mass-mailing platform (Ruby on Rails, AWS, Docker), completely automating HR workflows and cutting manual processing by 75%.

I also optimized the mobile media pipelines. By integrating FFmpeg video compression in Dart, I successfully reduced payload sizes by 50% and boosted low-bandwidth upload success rates by 40%.

JOMBAY // AWS_PIPELINE
Machine Learning Intern (Vision AI)
Center for Development of Advanced Computing (C-DAC, India)
Jun 2022 - Feb 2023

Mapping a country manually is an impossible task. We had terabytes of satellite imagery but no way to process it fast enough.

I deployed a U-Net architecture on HPC GPUs, effectively teaching the computer to "see" roads with 96% accuracy. This accelerated urban planning workflows by 70%.

C-DAC // U-NET_HPC

Projects

PEVAL // Multi-Agent Architecture
LangGraph & Qwen3 on Tau-Bench

LLM agents often get stuck in loops or lose context during complex tasks. To fix this on Tau-Bench, I developed the PEVAL LangGraph multi-agent framework. (Because an LLM that can't remember what it did 5 minutes ago isn't an agent, it's a goldfish.)

I built a Memory Kernel and a Global Learning Node. Benchmarking the Qwen3 family via vLLM on A100 GPUs, the 32B model's consistency metric shot up from 0.0% to 30.0%.

LangGraph vLLM Qwen3
PEVAL // MULTI-AGENT_FLOCK
VLM Tri-Agent Visual Navigation
GUI Automation on Intel Gaudi

GUI automation with Vision-Language Models often fails due to microscopic visual drift. You tell it to click a button, and it clicks empty space. (Which is fine until it accidentally clicks "Delete All" instead of "Save".)

I designed a Tri-Agent VLM pipeline on Intel Gaudi HPUs using Qwen3-VL, utilizing parallel coordinate clustering. This boosted OmniACT Action Scores to 51.5% and suppressed the AgentHARM Attack Success Rate to a mere 2.9%.

VLM Intel Gaudi HPUs PyTorch
VLM // GUI_TARGETING
DeepDebris // Space Domain Awareness
Multi-Agent RL & Neuro-Symbolic AI

Orbit is a chaotic minefield, and standard algorithms are too slow to survive it. I trained PPO Agents (Reinforcement Learning) to autonomously dodge debris, treating orbital mechanics like a high-stakes chess match.

I also built OrbitGPT, a RAG-based 'Space Lawyer' that cites international treaties in real-time. (Basically, I built a simulation to break things safely, because debugging in orbit is... expensive.)

Multi-Agent RL PPO RAG PPO
DEEP_DEBRIS // PPO_ORBITS
Connector-Aware Pretraining
Solving Gradient Starvation in LLMs

Standard LLMs obsess over nouns and starve the "connector" words (like "therefore"). It's like trying to build a bridge with only bricks and no mortar. (They sound confident, but they have the logical reasoning skills of a toddler.)

My fix? A surgical intervention on Llama 3.2. I identified 150+ discourse markers and amplified their gradients.

PyTorch Llama 3.2
LLAMA_3.2 // GRADIENT_PULSE
Precision Alignment
Qwen 2.5 Efficient Research

Using a massive 70B parameter model to judge a tiny 0.5B model is like using a sledgehammer to crack a walnut. It works, but it's wasteful. (And looking at the GPU bill made me physically ill.)

I pioneered a lightweight alignment pipeline for Qwen2.5 using a programmatic, heuristic reward function. 43.5% accuracy on GSM8K.

RLAIF Efficiency Alignment
QWEN2.5 // RLAIF_SYNC
Vision AI // Satellite Road Extraction
U-Net Architecture on HPC

Mapping a country manually is an impossible task. We had terabytes of satellite imagery but no way to process it fast enough.

I deployed a U-Net architecture on HPC GPUs, effectively teaching the computer to "see" roads with 96% accuracy. This accelerated urban planning workflows by 70%.

U-Net HPC Computer Vision
U-NET // KERNEL_SCAN

Transmission Feed

Synthesizing latest research and technical insights from Medium.

Initiating data uplink...

Tech Stack

AI/ML & Alignment
Transformers VLMs GNNs Diffusion Scaling Word2Vec RLHF RLVR PPO DPO TPO GRPO SFT
Agentic & Orchestration
MCP Multi-Agent Systems LangGraph LangChain Tool Calling ReAct Act OpenClaw Google ADK
RAG, Data & Benchmarks
Vector DBs Graph DBs Semantic Search Knowledge Graphs Tau-Bench Tau-Trait
Languages & Compute
Python C C++ SQL AWS Docker NVIDIA A100 GPUs Intel Gaudi HPUs TPUs

Ready to Connect?

Initiate a transmission. Let's build something extraordinary.