AI research. Systems engineering.
I'm a CS student at PES University, exploring LLM inference, AI/ML, and systems engineering. I enjoy turning research and ideas into software that actually works.
01 / About
I'm a final-year Computer Science student at PES University with a strong interest in AI/ML, systems engineering, and understanding how things work under the hood.
I enjoy diving deep into technical problems and turning what I learn into things that actually work. My current research focuses on efficient LLM inference, but I also love building software that solves real-world problems, from GPU-accelerated systems to financial analytics platforms.
Outside of tech, you'll usually find me playing tennis or the piano. I'm naturally curious and enjoy exploring new ideas, whether they're related to computer science or something completely different.
02 / Publications & ongoing work
Proposed a four-level taxonomy of hybrid LLM inference, identified the Orchestration Gap across inference levels, and introduced CPS and ITTS metrics for evaluating speed–quality tradeoffs.
Inference research · Ongoing
Developing a closed-loop LLM inference engine to address the Orchestration Gap through speculative decoding, domain-specialized routing, and conditional multi-model verification using shared real-time telemetry.
03 / Software & systems
LLM safety
Token-level LLM safety using two adversarially fine-tuned BadExpert LoRA adapters. A custom Hugging Face LogitsProcessor adjusts logit subtraction through a Representation Engineering router, evaluated on MMLU, TruthfulQA, and AdvBench.
Inference systems · Ongoing
Controlled Hierarchical Agent Speculative Engine. Designing adaptive routing that switches draft agents mid-generation using live acceptance signals and adjusts speculation length based on measured inference costs.
Repository will be made public soon.
Systems engineering
Multithreaded FastCDC chunking on the CPU, paired with CUDA SHA-256 hashing, double-buffered streams, and a crash-safe chunk store. Reached up to 5× throughput over a CPU-only baseline with byte-perfect restoration.
Financial market automation
A stock analytics platform with real-time monitoring, alerts, news ingestion, and portfolio analytics, supported by more than ten background services. Containerized with Docker Compose and deployed to an AWS VPS.
04 / Internships
June – August 2026
Translational AI for Networked Universal Healthcare, IISc
Built a multimodal mental health screening pipeline combining Whisper embeddings, ASR, metadata, and questionnaire features through early and late fusion. Developed speech processing across English, Hindi, Kannada, and Telugu using DeepFilter, Sarvam AI, and Gemini, and trained XGBoost, Logistic Regression, and SVM classifiers.
June – July 2022
Netcore Cloud
Built a production tool for automated API reliability monitoring, parsing large-scale logs to detect failed client requests.
05 / Academic background
August 2023 – July 2027 (expected)
B.Tech in Computer Science and Engineering
May 2021 – May 2023
IGCSE & A Levels