HariomGupta
AI Engineer and Researcher passionate about Generative AI, Multi-Agent Systems, LLM Applications, and Deep Learning. Dedicated to developing intelligent systems that convert advanced research into impactful real-world solutions.
Hariom Gupta
AI Engineer & Researcher
3+
Research Papers
10+
Projects Built
5+
ML Models Deployed
50+
Open Source Stars
Bridging Research & Engineering
I'm a Computer Science engineer passionate about applying deep learning and AI to solve meaningful real-world problems.
I hold a B.Tech in Computer Science & Engineering, where I discovered a deep passion for artificial intelligence and machine learning. What started as academic curiosity evolved into a commitment to building systems that genuinely matter.
My work lives at the intersection of research rigor and engineering pragmatism. I enjoy taking ideas from paper to production — whether that's a novel deep learning architecture for disease prediction or a transformer-based NLP pipeline processing millions of records.
I'm particularly drawn to healthcare AI — an area where intelligent systems can directly improve lives. I believe the best AI research doesn't just advance benchmarks; it translates into real-world tools that doctors, scientists, and everyday users can trust and rely on.
Outside of AI, I build full-stack applications, contribute to open source, write about ML engineering, and think deeply about the long-term trajectory of artificial intelligence.
Education
B.Tech, Computer Science & Engineering
Specialization: AI/ML & Data Science
2020 – 2024
Currently
Areas of Interest
Machine Learning
Designing and training models that learn from data to make intelligent decisions.
Deep Learning
Exploring neural architectures — transformers, CNNs, diffusion models, and beyond.
AI Research
Translating cutting-edge papers into working systems and publishing original findings.
Healthcare AI
Building intelligent diagnostic and predictive systems to improve patient outcomes.
Full Stack Dev
Creating polished, production-grade applications from ML backend to React frontend.
AI Systems
Designing scalable ML infrastructure — from model serving to distributed training.
Research & Publications
Exploring the frontier of AI/ML — from healthcare prediction systems to privacy-preserving learning.
Research Philosophy
I believe great research sits at the intersection of rigorous methodology and real-world applicability. My focus areas span healthcare AI, privacy-preserving ML, and neural architecture design — always aimed at systems that can be trusted and deployed.
Ask Early, Ask Late, Ask Right: When Does Clarification Timing Matter for Long-Horizon Agents?
TL;DR
Clarification timing matters for long-horizon agents — earlier isn't always better; optimal windows depend on information type and task structure.
This research examines when AI agents should request clarification during multi-step tasks. We develop a testing framework injecting clarifications at different execution points across four information types: goal, input, constraint, and context. Key findings reveal that the "earlier is always better" intuition is incorrect — goal clarification loses effectiveness quickly (pass@3 drops from 0.78 to baseline after 10% of execution) while input clarification remains valuable through approximately 50% completion. Analysis across 6,000+ experimental runs shows current frontier models rarely ask clarifications within optimal timing windows, and these timing patterns reflect inherent task characteristics rather than model-specific behaviors.
What I've Built
From research prototypes to production systems — here's a selection of work that reflects my engineering philosophy.
TPCA Leaderboard · needle
TokenBench
TokenBench is a measurement instrument for evaluating RAG, repo-map, graph, LLMLingua, and raw-dump context retrieval methods on real codebases. It introduces TPCA(V) — Tokens Per Correct Answer at amortization volume V — which correctly accounts for cold-start build costs vs. amortized query costs. Benchmarks run over two task families (needle-codebase & swe-qa) across three pinned Python repos (click, rich, httpx) with both AutoContains and LLM-judge scoring.
Technical Expertise
A snapshot of the tools, languages, and frameworks I work with daily — from research prototyping to production deployment.
Core Tools & Technologies
Machine Learning & AI
Experience & Timeline
From academic foundations to applied research and engineering — a timeline of growth and impact.
B.Tech in Computer Science & Engineering
KIIT University
- Specialized in Artificial Intelligence and Machine Learning
- Coursework: Deep Learning, Data Structures, Algorithms, Database Systems, Computer Networks
- Research focus on healthcare AI and predictive systems
- Completed major project on AI-powered disease prediction platform
AI Engineer Intern
PwC
- Six-month internship on the AI engineering team
- Contributed to production AI/ML systems and internal tooling
- Converted to full-time AI Engineer role at the end of the internship
AI Engineer
PwC
- Full-time AI Engineer building and shipping production AI systems
- Working across model development, evaluation, and deployment
Thoughts & Writing
Explorations in AI, machine learning engineering, research methodology, and the future of intelligent systems.
Let's Build Something
Open to research collaborations, ML engineering roles, consulting projects, and interesting conversations about AI.
What I'm open to
Let's push the frontier together
Whether it's a novel research idea, a hard engineering problem, or building the next generation of intelligent systems — I'm always up for an ambitious challenge.