Skip to content
View rishi-more-2003's full-sized avatar
🍃
🍃
16 followers · 14 following

Achievements

Achievement: YOLOAchievement: Pull SharkAchievement: Quickdraw

Achievements

Achievement: YOLOAchievement: Pull SharkAchievement: Quickdraw

Highlights

  • Pro

Block or report rishi-more-2003

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
rishi-more-2003/README.md

rishi

cs @ johns hopkins.
interested in efficient llm inference, rag, preference optimization, and ml systems.

co-first author on conformal thinking, accepted at icml 2026.

currently thinking about compute-aware reasoning, temporal rag, and data quality for post-training.

🔬 research

making llms lazy in a good way
conformal thinking — risk control for reasoning on a compute budget.
accepted at icml 2026.
arxiv

retrieval that remembers time
temporal-aware rag over non-stationary document collections.
time-aware indexing, retrieval freshness, and hallucination detection.

teaching a toxicity filter using reddit votes
preference optimization for context-dependent toxicity classification.

📄 publications

conformal thinking: risk control for reasoning on a compute budget
accepted at icml 2026 · arxiv

privacy-preserving video analytics through gan-based face de-identification
nmitcon 2024 · paper

toxic comment detection using bidirectional sequence classifiers
idciot 2024 · paper

harnessing insights from streams: unlocking real-time data flow with docker and cassandra in the apache ecosystem
raics 2024 · paper

prism: predictive resource inference and spot instance management
iconat 2024 · paper

a streamlined approach towards monkeypox detection
preprint · paper

preserving privacy in video analytics: a comprehensive review of face de-identification and background blurring techniques
preprint · paper

a lightweight approach towards speaker authentication systems
preprint / work in progress · paper

🛠️ stack

python · pytorch · tensorflow · hugging face · langchain · vllm
rag · inference optimization · preference optimization · calibration
docker · kubernetes · aws · spark · airflow
postgresql · mongodb · cassandra

💼 past work

jhu dsai
graduate research assistant — compute-aware reasoning and post-training data quality.

jhu clsp
graduate researcher — temporal rag and hallucination detection.

mastek
ml engineer intern — voice authentication and production inference.

🧊 elsewhere

website · google scholar · linkedin ·

Pinned Loading

  1. asym-rel-eff-kt asym-rel-eff-kt Public

    Investigating whether a persistent caregiver–child relationship improves knowledge transfer in language agents.

    Python

  2. post-training-failure-evals post-training-failure-evals Public

    Evaluation harness for detecting reward hacking, sycophancy, verbosity bias, and false-confidence failures in post-trained language models

    Python

  3. decode-roofline decode-roofline Public

    Kernel-level profiling of batch-1 decode on consumer GPU to prove that decode is memory-bandwidth-bound against the hardware roofline, then beating the baseline with a fused dequant+GEMV CUDA kerne…

    Python

  4. vllm-kernel-profiler vllm-kernel-profiler Public

    Kernel-level profiling suite for vLLM inference, focused on decode-time latency, KV-cache access patterns, Triton kernels, and CUDA performance metrics.

    Python

  5. ot-info-loss ot-info-loss Public

    Label-preserving domain adaptation with KL-regularized optimal transport.

    Python

  6. contextual-toxicity-dpo contextual-toxicity-dpo Public

    Preference optimization for context-aware toxicity detection using Reddit community feedback.

    Jupyter Notebook