Medaura - Agentic Pharmacy System
Medication errors are an information problem. The information exists - it is just not connected at the moment it matters.
FastAPILangGraphChromaDBLangfuseReact
Pranav DhiranAI Engineer · Researcher
Currently
- 01LFX’26 menteeHyperledger Cello
- 02Neurosymbolic SLMsPre-train + RL
- 03Agri-language modelKnowledge graphs
Everything I build started as a question I couldn't leave alone.
Everything started with a question I couldn't put down. I didn't pick a field so much as a habit — follow a question until it turns into something you can build, test, and hand to someone else. Some questions became models; some became tools; a few became stories worth telling.
Final-year B.Tech, Electronics & Telecom at SGGSIE&T Nanded. Tier-3 college, self-taught in most of what matters.
Recognition
- Twice took a team to the Smart India Hackathon national finals — 2024 and 2025.
- Top 6 globally at UWA Hack For Impact 2026.
Every role is a different vantage on the same question.
Jun 2026 – Present
LFX Mentee
Hyperledger Cello · Linux Foundation
Fabric has an operations problem: steep learning curve, verbose tooling, too much that shouldn't require an expert. I'm building an agent that collapses that — natural language in, Cello API call sequence out, operation executed.
- Designing an AI agent that takes natural language, reasons over Cello API call sequences, and executes operations — eliminating manual dashboard interaction
- The hard part isn't the LLM; it's knowing which API calls compose into what the user actually meant
Apr 2026 – Present
AI Research Intern
IRT, University of South Carolina
The bet: small models with symbolic constraints can do things large models can't — not in spite of their size, but because of it.
- Researching neurosymbolic SLM architecture and pre-training pipelines — integrating symbolic reasoning constraints into small language model training
- Working on RL-based fine-tuning (GRPO/RLHF) for SLM alignment — reward modeling, policy optimization, and evaluation on neurosymbolic reasoning benchmarks
Mar 2026 – Present
Open Source Contributor
Meshery — CNCF Sandbox Project
Five-plus merged PRs into a CNCF sandbox project. The PRs matter less than what you absorb reading other people's production code at scale.
- 5+ merged PRs — service mesh management features, UI components, and API integrations across Go backend and React frontend
- Active in code reviews, issue triage, and community discussions per CNCF contributor guidelines
Rabbit holes that became systems.
Small Language Model From Scratch - TinyStories
Every LLM course teaches you to call an API. I wanted to know what happens before the API.
PyTorchCustom BPEAMPNLP
03
GNU Radio MCP Server - LLM-to-SDR Bridge
LLMs can reason about RF signals. They just could not touch a radio. This closes that gap.
PythonFastMCPZMQXML-RPCGNU Radio
RF Watch - Open-Source Real-Time RF Spectrum Monitor
Physical-layer first: no protocol decoding, no black-box certainty, just traceable RF evidence.
PythonGNU RadioHackRF OneSignal Processing
The papers that gave me the vocabulary.
Not a reading list — each one changed what I thought was possible.
01arXiv
ReAct: Synergizing Reasoning and Acting
Agents observe before they act.
2210.0362902arXiv
Toolformer: Models Teach Themselves to Use Tools
The mental model for LLMs using external tools.
2302.0476103arXiv
Switch Transformers: Mixture of Experts
Modular capacity beats monolithic scaling.
2101.0396104arXiv
Group Relative Policy Optimization
A foundation for post-training interest.
2402.0330005arXiv
Direct Preference Optimization
Preference optimization without treating RL as magic.
2305.1829001arXiv
ReAct: Synergizing Reasoning and Acting
Agents observe before they act.
2210.0362902arXiv
Toolformer: Models Teach Themselves to Use Tools
The mental model for LLMs using external tools.
2302.0476103arXiv
Switch Transformers: Mixture of Experts
Modular capacity beats monolithic scaling.
2101.0396104arXiv
Group Relative Policy Optimization
A foundation for post-training interest.
2402.0330005arXiv
Direct Preference Optimization
Preference optimization without treating RL as magic.
2305.18290Thinking out loud — same act as building, different medium.
Writing02
Essays about training, alignment, and why I keep going back to first principles.
01
Why pre-train, not fine-tune?Fine-tuning patches behaviour. Pre-training shapes belief. One is a fix; the other is a foundation.
2025
02
How Do You Design a Custom SLM?From Qwen3-0.6B to a 133M agricultural language model — not shrinking, but reallocating the parameter budget where it counts.
Aug 2026
Speaking02
Gave lectures at IAIRO SLM++ Bootcamp (2026) — Vanilla GPT-2 Architecture & Evolution of LLM Design Decisions — Scaling Laws, Cost Accounting & Case Studies.
If any of this resonated, let's talk.
Currently interested in AI research internships, open-source collaborations, and systems engineering opportunities. Cold emails work.