Principal InvestigatorPlasma-tissue interaction, SJRIC funded
Patents pendingAlena AI tutor · chaotic encryptor
AuthorSimplified Computational Physics, 2025
Phase space · liveTap to inject electronsv / vth ↑x / λD →
trapped 0 · passing 0
Test electrons in a fixed electron-hole potential, H = v²/2 − φ(x). Gold particles are trapped, blue are passing. Illustrative; not a self-consistent Vlasov-Poisson solution.
Research · 1
Plasma kinetic theory
Vlasov-Poisson dynamics of collisionless plasmas. Doctoral work, continued with PhD and Master's students.
7published2accepted
Electron and ion holes are self-trapped vortices in phase space. I study how they form, accelerate and merge, and what a fluid description of phase space reveals about them.
Lobo, A.* & Sayal, V. K. · Physics of Plasmas 31(9), 092301
Published
2026
Analytical and numerical investigation of soliton collisions in the Schamel-Korteweg-de Vries limit
Sen, M., John, D., Lobo, A.*, Saravanan, A. · Springer Proceedings in Physics (ICNDA 2026)
Accepted
2026
Numerical investigation of Hall thruster discharge in plasma
Dev, N., Sen, M., Raghv, T., Lobo, A.*, Saravanan, A. · Springer Proceedings in Physics (ICNDA 2026)
Accepted
Research · 2
Probability theory and active Brownian entanglement
Feedback-driven probability, and correlations in active Brownian systems.
3published1accepted1in review
My contribution to foundational probability is a feedback-driven account of how repeated random events evolve and converge.
Each outcome feeds back into the chance of the next. Weak feedback recovers ordinary convergence; strong feedback locks runs into different limits. Move the slider to see the change.
Repeated trials · 14 runsfrequency ↑trial n (log) →
spread at n = 600: 0
Illustrative reinforcement model, pₙ = ½ + k (fₙ − ½), where fₙ is the running frequency. Not the published formulation.
I also study classical entanglement in active Brownian particles: how activity induces it, and how system size limits it.
Self-propelled particles in a trap do not settle at the centre. As activity rises they pile up at a rim set by their speed, and their motion becomes persistent and correlated.
Active particles in a harmonic trap
mean radius 0
Illustrative non-interacting active Brownian particles. The dashed circle marks r = v₀/κ. The published work adds interactions and computes an entanglement witness over ensembles.
Saravanan, A. & Lobo, A. · Physical Review E 112(2), 024138
Published
2026
Physical control of classical entanglement in interacting Brownian particles
Kashyap, P., Neeraj, B., Lobo, A., Saravanan, A.* · Springer Proceedings in Physics (ICNDA 2026)
Accepted
2026
Activity-induced Brownian entanglement in trapped active particles and its finite-size limit
Lobo, A.*
In review
Research · 3
Computational biophysics
Molecular dynamics, steered MD and continuum electrostatics on proteins in their physiological setting.
1funded PI1preprint1award
The same questions about transport under driving, asked of proteins: how ligands cross membranes, how motors choose a direction, and what reactive species from cold plasma do to living tissue.
The kinesin work rests on a simple idea you can run here. Switch a lopsided potential on and off, and random thermal motion turns into steady drift. Turn the flashing off and the drift stops.
Flashing Brownian ratchetpotential on
mean drift 0.00 periods
Illustrative overdamped particles in an asymmetric sawtooth potential of height 5 kBT. The kinesin model itself uses APBS electrostatics and steered MD on the real protein.
Mechanisms and dynamics of plasma-tissue interaction
Analytical and computational models of how cold atmospheric plasma acts on biochemical structures, focused on ligand transport and metabolic activation driven by reactive oxygen and nitrogen species (RONS), with a view to plasma therapy in cancer.
Funded project 2026RDCSJRI031 · St Joseph's Research and Innovation Centre · INR 4.5 lakh · ongoing
MethodCHARMM-GUI systems, GROMACS MD, CP2K
QuestionHow RONS alter transport kinetics and catalytic activation
RolePrincipal Investigator
Kinesin-1 as an electrostatically biased Brownian ratchet
APBS continuum electrostatics on a five-dimer protofilament and steered MD across apo, ADP and AMPPNP states with Jarzynski analysis. Fixed-field runs show a dominant 41 Å symmetric mode rather than the 82 Å directional step, supporting a biased ratchet over a deterministic power stroke.
In review · On the electrostatic and stochastic physiology of kinesin protein dynamics · Lobo, A.* (2026)
Finding41 Å symmetric mode dominates in fixed-field runs
TalkProtein electrostatics and stochasticity, ICAR-CMFRI, Kochi (2026)
Glucose translocation through GLUT4
Thermodynamic and kinetic profiling of glucose transport in POPC-embedded GLUT4, comparing the cryo-EM structure with AlphaFold models across apo, ATP and ADP states using steered MD.
bioRxiv preprint ↗ · Govinda S., Das S., Lobo, A.*, Rahul C. N., Gideon D. A. (2026)
MethodMembrane-embedded MD and steered MD in GROMACS
ComparisonCryo-EM structure versus AlphaFold models
TalkSmartChem 2026, St Joseph's University
pH-dependent destabilisation of CD36
In-silico modelling of how the acidic tumour microenvironment destabilises the CD36 fatty-acid transporter, with implications for lipid uptake in oral squamous cell carcinoma.
2nd Prize · SLMDA Spring Scientific Session, London (2026) · Lobo, A., Jayasooriya, P., De Silva, G., Perera, S. M. W.
MethodProtonation-state MD of CD36 at physiological and tumour pH
ContextFatty-acid uptake in the tumour microenvironment
Recognition2nd Prize, paper presentation, London
Engineering · 4
Developer and AI engineer
Privacy-first AI platforms, physics-informed machine learning and production tools, designed and built end to end.
2patents pending1copyright
I build machine learning as structured dynamical systems with causality, feedback and conservation built in, so the models stay interpretable. The platforms below put that approach and retrieval-augmented generation (RAG) into working products.
Academic AIEngine behind Alena
Noel AI: a two-brain academic assistant
Answers from the institution's own material and keeps documents on the machine. Retrieval and generation are separate brains with a fixed privacy boundary between them.
Grounded, source-bound answers designed to resist hallucination
Code mode for MATLAB and Python, with sandboxed Python runs
Teaching mode with rendered LaTeX equations
Apple Silicon and Intel Macs, Windows, Ubuntu
Python
FastAPI · uvicorn
React · Vite
sentence-transformers
Ollama
Groq
Patent pendingMobile
Alena: mobile AI tutor
The student-facing side of Noel AI. A self-learning and tutoring app whose answers come from course PDFs the teacher uploads, with an online and offline research assistant mode for faculty.
Mobile
RAG
Noel AI engine
CopyrightEd-tech platform
CLupus: AI-embedded LMS and ERP
A full learning management system for colleges: courses and materials, assignments and grading, quizzes and exams, attendance and timetables. Every course has Alena built in as its AI tutor.
Web
AWS Lightsail
Alena integration
ML control systemNegotiation engine
Nonlinear matrix control for negotiation
Negotiation as two co-evolving agents. The salesperson is a small neural network controlling a focus vector and a rank-2 reframing matrix; the customer carries a priority vector and a resistance-coupling matrix for linked objections. Purchase probability follows from the push and pull of the two.
Sir Sales-a-lot: live negotiation agent with attribute generation, a strict price floor and dynamic asking price
Trained against stingy, rational and impulsive customer archetypes
Deal probability by turn
Illustrative trajectories from a reduced scalar push-resistance model, P = σ(push − resistance). The engine itself uses full rank-2 matrices.
PyTorch
Streamlit
Groq · Llama 3.3 70B
Jupyter
Variational optimisationEffort estimator
Effort estimator and optimiser
An Euler-Lagrange derivation shows that optimal strategy pacing follows a catenary. The same rank-2 framework transfers one-to-one from negotiation to software workflow, giving effort estimates and an optimised pacing curve for a project.
Lower tension front-loads and back-loads effort; higher tension flattens it toward a uniform pace.
Effort rate across the project
start : midpoint = 0
e(t) ∝ cosh((t − ½)/a), normalised to the same total effort as a uniform pace (dashed).
Enterprise AIConsultant architect
Enterprise proposal platform
An offline platform that reads RFPs and historical project data to generate proposals, estimate effort, plan resources and support sales, with the negotiation and effort engines inside. Data stays on the company's machines.
FastAPI
ChromaDB
SSE streaming
Ollama · MLX
Interpretable ML
Dynamical neural networks
Networks built as dynamical systems with physical constraints such as causality, feedback and conservation, so the learned behaviour can be read and checked. Used across the projects above.
TensorFlow
PyTorch
MATLAB
Content automation
ShortMaker
A local pipeline that turns one prompt into a finished short vertical video for any online video platform.
An encryption scheme keyed by the chaotic dynamics of a double pendulum, where sensitivity to initial conditions does the work of a key schedule.
Nonlinear dynamics
Python
Academic · 5
Teaching, talks and service
Thirteen years of physics teaching, from entrance-exam coaching to postgraduate coordination.
Appointments
2025 – nowAssistant Professor of Physics, St Joseph's UniversityPG Programme Coordinator · PhD Coordinator · OBE Convenor · University AI Cell Coordinator
2026Enterprise Architecture Advisor (consultant)
2022 – 2024Teaching Assistant, Sikkim Manipal Institute of Technology
2021Assistant Professor, The Oxford College of Science
2018 – 2021Physics Lecturer, GATE-IIT Bangalore
Teaching and supervision
Classical and Quantum Mechanics, Thermodynamics, Numerical Techniques, Plasma Physics. Supervising one PhD student on kinetic studies of solitary and shock structures in plasmas, and seven Master's dissertations across plasma, biophysics and quantum computing.
Selected talks
2026 · LondonIn-silico modelling of the pH-dependent destabilisation of CD362nd PrizeSLMDA Spring Scientific Session
2026 · KochiProtein electrostatics and stochasticityInvited, ICAR-CMFRI national seminar
2026 · ChintamaniComputational techniques from biophysics to quantum computingInvited, Govt First Grade College & PG Centre
2026 · BengaluruGlucose translocation in POPC-embedded GLUT4SmartChem 2026, SJU
2025 · TirupatiSoliton propagation, collision and coalescence in the Schamel-KdV limitPlasma 2025, IIT Tirupati
2023 · DelhiElectron-hole coalescence in the phase-space fluid modelTPS 2023, JNU
Book and memberships
Simplified Computational Physics (2025), ISBN 979-8275373783. Member of the American Physical Society, AAPPS Division of Plasma Physics, Plasma Science Society of India and Indian Association of Physics Teachers. Best Paper Award, Sikkim Manipal University (2022).
Contact
Get in touch
Research collaboration, invited talks and platform demonstrations.
