I build the infrastructure layer that makes LLM and AI-agent systems run in production — reliably, on serverless/decentralized GPUs, at a fraction of managed-API cost. Published ML researcher (7 peer-reviewed papers, 4 in Q1 journals, IF up to 7.1) who ships the hard part: multi-agent orchestration, RAG, and GPU deployment — not just notebooks.
- 🚀 Production LLM / agent deployment — RunPod · Modal · Vast.ai; Dockerized serverless endpoints with CI/CD, plus self-contained installers that survive real GPU-provider failures (driver mismatches, multi-GPU OOM, topology-aware placement)
- 🧠 Multi-agent systems with latent-space collaboration — agents exchange hidden states instead of generated text → fewer tokens, faster inference (open-sourced as LatentMAS-SLoRA)
- 📚 RAG / document intelligence — embedding retrieval, citation tracking, multi-hop reasoning, and LLM-as-judge evaluation
- 🎛️ Dynamic LoRA serving — hot-swappable, domain-routed PEFT adapters on a single base model, with custom LoRA training built in
- 💸 Inference cost optimization — meaningful GPU savings and ~$0 idle cost via serverless autoscaling
📨 Tell me your model and your latency / cost target — I reply within 2 hours.
|
Light-Heart-Labs / DreamServer — a production, self-hostable AI-deployment platform for decentralized/serverless GPUs. My contributions are merged in upstream (16 PRs, 13 merged) and span the hard infrastructure surface:
Stack: Bash installer hardening · Docker Compose · GPU tier detection · FastAPI · pytest · ShellCheck · Vast.ai quirks |
|
Officially featured as a community extension in Gen-Verse / LatentMAS — the repo for the LatentMAS paper (arXiv:2511.20639, 🤗 #1 Paper of the Day). A multi-agent reasoning framework that augments LatentMAS with role-specialized, dynamically switchable LoRA adapters and latent-space collaboration — agents communicate through hidden states, not generated text. Includes VLM support (Qwen2.5-VL-7B), RAG integration, real evals, and RunPod serverless deployment with CI/CD. Measured results: +12% accuracy · 2.7× faster inference · 63.6% token reduction vs. traditional RAG.
|
|
LLM / Agent Infra
ML / DL
MLOps / Infra
Languages & Data
|
Multi-agent reasoning with dynamic LoRA routing + latent-space collaboration. Featured in Gen-Verse/LatentMAS.
|
Retrieval pipeline over URL/JSON/CSV docs with citation tracking and automated LLM-as-judge evaluation.
|
|
Serverless RunPod worker for FLUX.2 Klein text-to-image / image-to-image — $0 idle, autoscaled.
|
End-to-end serverless inference ecosystem for diffusion models with Dockerized endpoints + CI/CD.
|
|
Topology-aware multi-GPU pipeline with per-GPU VRAM capping (multi-GPU OOM fix) and self-host UI.
|
Self-hosted Flux LoRA platform tuned for A40 (48 GB) → H200, with 4-bit quantization & batch inference.
|
|
Serverless speech-to-text worker — scalable transcription endpoint on demand.
|
Low-cost, CPU-friendly self-hosting kit for local AI tooling with secure external sharing.
|
More projects — LoRA training, NLP fine-tuning & low-VRAM inference
| Project | What it does | Tech |
|---|---|---|
| Lora_Trainer_Imgen_Flux | Custom LoRA training pipeline for Flux image models | PEFT Flux |
| bangladeshi-tts-finetuning | Fine-tuning TTS for Bangla speech | TTS Fine-tuning |
| bangla-punctuation-restoration | Transformer-based punctuation restoration for Bangla | NLP Transformers |
| time-series forecasting (SOTA) | State-of-the-art time-series forecasting models | PyTorch Forecasting |
| Wan2GP | Low-VRAM video generation for consumer GPUs | PyTorch Gradio |
| Ltx-Image-to-Video | Image-to-video generation with a Gradio interface | LTX Gradio |
7 peer-reviewed papers · 4 in Q1 SCI journals (IF up to 7.1) · Google Scholar · ORCID · ResearchGate
| # | Paper | Venue | IF · Quartile | Year | Link |
|---|---|---|---|---|---|
| 1 | Machine-learning-assisted revelation of the best-performing single heterojunction thermophotovoltaic cell | Sustainable Energy Technologies & Assessments | 7.1 · Q1 | 2025 | DOI · arXiv |
| 2 | Machine-learning-enabled performance exploration of AuCuSe₄ in a thermophotovoltaic cell | Solar Energy | 6.0 · Q1 | 2024 | DOI |
| 3 | Numerical studies on a ternary AgInTe₂ chalcopyrite thin-film solar cell (first author) | Heliyon | 4.0 · Q1 | 2023 | DOI |
| 4 | Numerical prediction of the photovoltaic performance of a CZTS-based thin-film solar cell | Nano Select | Q1 | 2023 | DOI |
| 5 | Unleashing the power of open-source transformers in medical imaging (brain MRI, ~99.6% acc.) | IJACSA | — | 2024 | DOI |
| 6 | Spectrum estimation for voiced speech using average weighted linear prediction | Speech processing | — | 2024 | Profile |
| 7 | Enhancement of bone-conducted speech using deep transfer learning | Speech processing | — | 2024 | Profile |
Research themes: applied ML for materials/PV optimization, medical-imaging transformers, and speech enhancement.
- 🎓 B.Sc. Electrical & Electronic Engineering — University of Rajshahi, Bangladesh
- 🏅 Deep Learning with TensorFlow (IBM) · Prompt Engineering (Vanderbilt) · SQL Advanced (HackerRank)
