class SohamJadhav:
def __init__(self):
self.name = "Soham Jadhav"
self.role = "AI Engineer"
self.education = "B.E. AI & Data Science @ MET's Institute of Engineering, Nashik (2022–2026)"
self.focus = ["Deep Learning", "Generative AI", "Agentic Workflows", "Computer Vision"]
self.gsoc = "GSoC 2026 Applicant → ML4Sci DeepLense (DEEPLENSE6, DEEPLENSE7, PREDICT1)"
self.open_source = "pgmpy contributor"
self.building = ["Copilot for Data Science", "Multimodal Org Communication System"]
self.email = "soham.ai.engineer@gmail.com"
def current_mission(self):
return "Ship AI that actually works. No fluff."
|
|
|
|
|
|
|
|
| 🎓 Degree | 🏛️ Institution | 📅 Year |
|---|---|---|
| B.E. — Artificial Intelligence & Data Science | MET's Institute of Engineering, Nashik | 2022 – 2026 |
| Certificate | Issuer | Verify |
|---|---|---|
| 🏅 AI Engineering Professional | IBM / Coursera |  |
| 🏅 Deep Learning with PyTorch, Keras & TensorFlow | IBM / Coursera | |
| 🏅 Generative AI Engineering with LLMs | IBM / Coursera | |
| 🏅 Career Essentials in Generative AI | Microsoft & LinkedIn |  |
| 🏅 Python | HackerRank |  |
| 🏅 SQL Advanced | HackerRank | |
| 🏅 Co-Lead — AI & Machine Learning | Google Developer Groups |  |
Co-Lead — AI & Machine Learning · Google Developer Groups (GDG) On Campus, Nashik (Sept 2024 – Oct 2025) Taught and mentored peers in AI, ML, and Generative AI. Organized hands-on sessions on emerging technologies.
Convo-Ease: Intelligent Multi-Modal Moderation for Digital Organizational Communication
| Field | Details |
|---|---|
| 👥 Authors | Soham S. Jadhav, Nisha D. Patil, Omkar N. Gadakh, Atharv S. Gaikwad |
| 🏛️ Institution | MET's Institute of Engineering, Nashik |
| 📰 Publisher | Cureus — Part of Springer Nature |
| 📅 Year | 2025 |
| 🏷️ Keywords | AI Moderation · Multimodal Learning · Text-Image-Audio Fusion · Policy-as-Prompt |
Abstract: Proposes Convo-Ease — an AI-driven moderation framework using a Gatekeeper Architecture that validates content pre-delivery across text, image, and audio modalities in organizational chat systems. Introduces a Policy-as-Prompt method achieving sub-3-second latency with Gemma 3, BLIP-2, and Whisper.
Beyond Text: A Comprehensive Survey of Multimodal Content Moderation Architectures in Enterprise Environments
| Field | Details |
|---|---|
| 👥 Authors | Soham S. Jadhav, Omkar N. Gadakh, Nisha D. Patil, Atharv S. Gaikwad |
| 🏛️ Institution | MET's Institute of Engineering, Nashik |
| 📰 Publisher | ICIA Conference Proceedings |
| 📅 Year | 2025 |
| 🏷️ Keywords | Multimodal Moderation · LLMs · Gatekeeper Architecture · Enterprise Security · Audio-Visual Fusion |
Abstract: Surveys 26+ content moderation methodologies ranging from LLM-based guardrails to multimodal fusion architectures. Critically examines the transition from reactive API-based moderation to dynamic Policy-as-Prompt frameworks, identifying critical research gaps in latency management and on-premise privacy preservation.
- 🔭 Get accepted into GSoC 2026 with ML4Sci DeepLense
- 🤝 Merge open-source contribution into pgmpy
- 🚀 Launch Copilot for Data Science as a standalone product
- 📝 Publish first technical writeup / research note
- 💼 Land a full-time AI/ML Engineer role
# ──────────────────────────────────────────────
# T H E E N G I N E E R ' S M A N T R A
# ──────────────────────────────────────────────
from mindset import Curiosity, Discipline, Patience
from reality import Problems, Impact
def path_to_success(dream: str) -> Impact:
goal = {
"start" : "where_you_are_now",
"aim" : dream,
"fuel" : Discipline + Curiosity
}
for iteration in range(0, ∞): # no finish line
step_1 = research(problem_deeply) # understand before solving
step_2 = learn(tools_to_solve_it) # sharpen before striking
step_3 = build(something_real) # ideas only count shipped
if iteration % 100 == 0:
reflect() # recalibrate, not retreat
# "The expert in anything was once a beginner."
return Impact(measurable=True, lasting=True)"The goal is not to build AI. The goal is to build AI that matters."