Ashish Kumar

AI Researcher

University Park, Pennsylvania, United States6 yrs 10 mos experience

AI EnabledHighly Stable

Key Highlights

Expert in Large Language Models and Automated Prompt Optimization.
Strong background in distributed systems and software development.
Proficient in multiple programming languages and machine learning techniques.

Stackforce AI infers this person is a highly skilled AI and Software Engineering professional with expertise in distributed systems.

Contact

Skills

Core Skills

Large Language Models (llm)Code GenerationDistributed SystemsSoftware Development

Other Skills

LaTeXObject-Oriented Programming (OOP)JavaC#Agentic SystemsData AnalyticsJavaScriptData ScienceData StructuresDeep Reinforcement LearningReinforcement LearningNatural Language Processing (NLP)PyTorchJAXComputer Vision

About

My research interests include Large Language Models - specifically in Large Multimodal Models (LMMs), Post Training (SFT, RLHF, RLVR, RLAIF), Mechanistic Interpretability, Personalization of LLMs, Agentic AI, RAG systems, Language Diffusion Models, Prompt Optimization, Optimization Algorithms. I am also interested in classical Machine Learning Research on deep neural networks, CNNs, RNNs, LSTM, GRUs, recommender systems, regression, decision trees, clustering. I have also conducted research in Programming Languages, Graph Theory, Complexity Theory and Algorithmic Game Theory. I am proficient with python, numpy, scipy, sklearn, pytorch, jax, tensorflow, sympy, pandas. Also proficient in other languages like C, C++, Java, C#, Rust, OCaml, Coq, HTML, Javascript.

Experience

6 yrs 10 mos

Total Experience

6 yrs 10 mos

Average Tenure

6 yrs 10 mos

Current Experience

Google

Phd Intern

May 2025 – Aug 2025 · 3 mos · California, United States · On-site

1. Patent Pending...
In addition to Patent,
2. Implemented an engine for Automated Prompt Optimization specifically Hard Prompt Optimization for LLMs.
3. Implemented an engine capable of -
A multi-turn LLM Agent for information extraction.
LLM based automated code generation
Dashboard Consolidation and Search
4. Gave 2 internal talks surveying latest research in Automated Hard Prompt Optimization. Presented internship to CCA and OFDF teams.

Code GenerationLarge Language Models (LLM)

Amazon

Software Engineer Internship

May 2024 – Aug 2024 · 3 mos · San Jose, California, United States · On-site

1. Contributed to extending the P framework (link: https://p-org.github.io/P/whatisP/) for distributed systems by exploring the search space of all possible schedules to identify buggy schedules or verify correctness.
2. Worked on horizontally scaling the search space exploration strategy by using multiple threads to explore the search state.
3. Worked with various search space techniques, including random walks, DFS tree traversal, and the A-star search algorithm.
4. Explored ideas for machine-level parallelism to enhance scaling beyond thread-level (core-level) parallelism.
5. Presented future research directions, including:
(i) Enhanced search space exploration using biased random walks, with probabilities based on metrics tailored for distributed protocols.
(ii) Improved parallelism speedup through better partitioning of work among threads.
(iii) Symbolic execution techniques to reduce search space size effectively.

JavaC#Distributed SystemsSoftware Development